University of Groningen Credit supply and macroeconomic … · 2018. 3. 12. · 517798-L...
Transcript of University of Groningen Credit supply and macroeconomic … · 2018. 3. 12. · 517798-L...
University of Groningen
Credit supply and macroeconomic fluctuationsPool, Sebastiaan
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.
Document VersionPublisher's PDF, also known as Version of record
Publication date:2018
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):Pool, S. (2018). Credit supply and macroeconomic fluctuations. University of Groningen, SOM researchschool.
CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.
Download date: 22-08-2021
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 1PDF page: 1PDF page: 1PDF page: 1
Credit supply and macroeconomicfluctuations
Sebastiaan Pool
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 2PDF page: 2PDF page: 2PDF page: 2
Publisher: University of Groningen, Groningen, The Netherlands
Printed by: Ipskamp Printing
P.O. Box 333
7500 AH Enschede
The Netherlands
ISBN: 978-94-034-0525-4 / 978-94-034-0524-7 (ebook)
c© 2018 Sebastiaan Pool
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system of any nature, or transmitted in any form or by any means,
electronic, mechanical, now known or hereafter invented, including photo-
copying or recording, without prior written permission of the publisher.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 3PDF page: 3PDF page: 3PDF page: 3
Credit supply and macroeconomicfluctuations
PhD Thesis
to obtain the degree of PhD at the
University of Groningen
on the authority of the
Rector Magnificus, Prof. E. Sterken
and in accordance with
the decision by the College of Deans.
This thesis will be defended in public on
Thursday 22 March 2018 at 16:15 hours
by
Sebastiaan Pool
born on 26 December 1990in Deventer, The Netherlands
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 4PDF page: 4PDF page: 4PDF page: 4
Supervisor
Prof. J.M. Berk
Co-supervisor
Dr. J.P.A.M. Jacobs
Assessment committee
Prof. B.J. Heijdra
Prof. C.G. de Vries
Prof. F.R. Smets
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 5PDF page: 5PDF page: 5PDF page: 5
Acknowledgements
The past three years I have spent the majority of my time writing this disser-
tation. Physically I was located primarily at De Nederlandsche Bank where
I had the opportunity to glance into a tumultuous policy environment. I
remember the countless conversations with colleagues discussing contem-
porary economic issues which have not only benefited the quality of my
research, but also shaped me into a more general economist. Meanwhile I
was always able to retrieve from the fads and enjoy the tranquility of the
University of Groningen. Working for both institutions was a true pleasure
and felt like a continuation of my studies. These memories characterize the
professional, but also collegial environment I worked in. For this I am truly
thankful.
Many people in particular have contributed to this research and I would
like to express my gratitude to them. First of all, I would like to thank my
supervisor, Jan Marc Berk, for having confidence in me from the start. Your
economic insights and critical assessment on my writings have been invalu-
able for this work. Second, I would like to thank my co-supervisor Jan Jacobs
for your constructive comments on the econometric tests, but also for shar-
ing your experience in writing an academic paper. Third, I would like to
extend my thanks to my co-authors; your contribution has taken this dis-
sertation to a higher level. Niels Gilbert, thank you for sharing your policy
experience and writing skills with me. Leo de Haan, your pragmatism in
both research and policy was illuminating.
Moreover, I would like to thank a few others who have helped me to
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 6PDF page: 6PDF page: 6PDF page: 6
ii Acknowledgement
write this dissertation. Jakob de Haan for offering me the opportunity to
write this dissertation and for commenting on my work repeatedly. Peter
van Els, not only for sharing your relentless enthusiasm for the economic
profession, but also for giving me constructive feedback and for taking the
heat off me occasionally. Mark Mink, for being a sounding board group
and for giving (un)solicited advice about everything, but mostly about our
shared passion: economics. Renske Maas for sharing thoughts on economics
and sports, but also for our mental reboot (coffee) twice a day.
I am truly grateful to my parents who have given me the opportunity
and provided me with the means to grow both socially and professionally.
Finally, I would like to thank Ellen Schipper for discussing anything but
economics and for unconditional support.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 7PDF page: 7PDF page: 7PDF page: 7
Contents
Acknowledgements i
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Aim of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Loan loss provisioning, bank credit and the real economy 11
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.1 Macroeconomic framework . . . . . . . . . . . . . . . . 15
2.2.2 What does a provisioning shock do? . . . . . . . . . . . 20
2.2.3 Empirical setup . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4.1 Main results . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.2 Robustness . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.A Model solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.B Variable names and definitions . . . . . . . . . . . . . . . . . . 35
2.C Robustness checks . . . . . . . . . . . . . . . . . . . . . . . . . 35
3 Credit defaults and bank capital 39
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 8PDF page: 8PDF page: 8PDF page: 8
Contents
3.2 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2.1 The real side: households, entrepreneurs and firms . . 43
3.2.2 The financial side: banks . . . . . . . . . . . . . . . . . . 50
3.2.3 Bank recapitalizations and countercyclical buffers . . . 59
3.3 Empirical methodology . . . . . . . . . . . . . . . . . . . . . . 63
3.3.1 Calibrated parameters . . . . . . . . . . . . . . . . . . . 64
3.3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3.3 Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.4 Empirical results . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.4.1 Technology shock . . . . . . . . . . . . . . . . . . . . . . 72
3.4.2 Credit default shock . . . . . . . . . . . . . . . . . . . . 75
3.4.3 Countercyclical capital buffer . . . . . . . . . . . . . . . 81
3.4.4 Endogenous recapitalization . . . . . . . . . . . . . . . 84
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.A Model solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.B Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4 Mortgage loans and shadow banks 105
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.2 Stylized facts extended . . . . . . . . . . . . . . . . . . . . . . . 113
4.2.1 Linear regression model . . . . . . . . . . . . . . . . . . 113
4.2.2 Multivariate regression model . . . . . . . . . . . . . . 115
4.3 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.3.1 Aggregate risk . . . . . . . . . . . . . . . . . . . . . . . 121
4.3.2 Real economy: households and firms . . . . . . . . . . 122
4.3.3 Regulated and shadow banks . . . . . . . . . . . . . . . 127
4.3.4 How do banks invest? . . . . . . . . . . . . . . . . . . . 130
4.3.5 Expected liquidation value . . . . . . . . . . . . . . . . 131
4.3.6 Externalities . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.3.7 Closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.3.8 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 9PDF page: 9PDF page: 9PDF page: 9
Contents
4.4.1 Growth of mortgage loans . . . . . . . . . . . . . . . . . 139
4.4.2 Shadow bank growth . . . . . . . . . . . . . . . . . . . 143
4.4.3 Liquidity risk and the lender of last resort . . . . . . . . 145
4.5 Policy options . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
4.5.1 Loan-to-value constraints . . . . . . . . . . . . . . . . . 147
4.5.2 Interest on cash . . . . . . . . . . . . . . . . . . . . . . . 149
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.A Variable names and definitions . . . . . . . . . . . . . . . . . . 153
4.B Household and firm problem . . . . . . . . . . . . . . . . . . . 155
4.C Bank optimization problem . . . . . . . . . . . . . . . . . . . . 158
4.D Proof Proposition 4.1 . . . . . . . . . . . . . . . . . . . . . . . . 161
4.E Including idiosyncratic credit risk . . . . . . . . . . . . . . . . . 162
5 Sectoral allocation and macroeconomic imbalances in EMU 165
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.2 Stylized facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.3 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.3.1 Households . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.3.2 Firms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
5.3.3 Monetary authority and government sector . . . . . . . 178
5.3.4 Market equilibrium conditions . . . . . . . . . . . . . . 179
5.3.5 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.4 Model simulations . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.4.1 Two-region model . . . . . . . . . . . . . . . . . . . . . 182
5.4.2 Including the Rest of the World . . . . . . . . . . . . . . 186
5.4.3 Crisis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.5 Empirical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.5.1 Methodology and data . . . . . . . . . . . . . . . . . . . 187
5.5.2 Empirical results . . . . . . . . . . . . . . . . . . . . . . 191
5.6 Policy options and discussion . . . . . . . . . . . . . . . . . . . 193
5.6.1 Increasing competition in the nontradable sector . . . . 194
5.6.2 Deepening the internal market . . . . . . . . . . . . . . 195
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 10PDF page: 10PDF page: 10PDF page: 10
Contents
5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
5.A Sectoral dependence on domestic demand . . . . . . . . . . . . 198
5.B Model solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
5.C Including the Rest of the World . . . . . . . . . . . . . . . . . . 204
5.D Robustness checks . . . . . . . . . . . . . . . . . . . . . . . . . 206
6 Conclusion 211
6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.2 Policy implications . . . . . . . . . . . . . . . . . . . . . . . . . 213
6.3 Future research . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
Summary 233
Samenvatting (summary in Dutch) 237
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 11PDF page: 11PDF page: 11PDF page: 11
Chapter 1
Introduction
1.1 Background
In the decades preceding the global financial crisis, the macroeconomic lit-
erature largely ignored the role of credit supply in determining macroe-
conomic fluctuations. This limited role for credit supply stands in sharp
contrast to early seminal work in the field. In the view of, e.g., Thornton
(1802), Wicksell (1898) and Fisher (1933, 1961) macroeconomic fluctuations
are, above all, caused by the expansion and contraction of credit. Similarly,
Keynes (1930, 1936) argues that credit supply, which is (in part) determ-
ined by the confidence of lenders to finance borrowers, is an important
factor in determining investment and output fluctuations. In the decades
that followed, however, the importance of fluctuations in credit supply in
the macroeconomic literature declined considerably.1 Consequently, the role
of credit supply was largely absent in the dominant class of macromodels,
the so-called Dynamic Stochastic General Equilibrium (DSGE) models, that
were developed at the turn of the millennium and widely used.2
The development of DSGE models without a central role for credit sup-
ply was supported by theory. In particular, abstracting from credit supply in
1 Economists that continued to focus on the role of credit supply in determining macroe-conomic fluctuations, most notably Minsky (1986), were no longer considered to be main-stream.
2 See e.g. Christiano et al. (2005) and Smets and Wouters (2007) for two well-known work-horse DSGE models.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 12PDF page: 12PDF page: 12PDF page: 12
2 Chapter 1
the frictionless general equilibrium framework of Arrow and Debreu (1954)
has no effect on real or nominal variables. First, Modigliani and Miller (1958)
formally showed that financial structure—the proportion of debt and equity
finance—is both indeterminate and irrelevant for real economic outcomes.
As a result, in the frictionless general equilibrium framework of Arrow and
Debreu banks could be omitted because whether representative households
finance representative firms directly or through banks is irrelevant for the
determination of real variables.
A few years later, Gurley and Shaw (1960) studied the role of financial
markets and institutions in the real economy. They argued that in a neo-
classical setting the presence of financial markets and institutions does not
restrain the ability of central banks to determine the price level. It is, how-
ever, crucial to assume demand for central-bank liabilities.3 Conditional on
demand for its liabilities, the central bank can determine and manipulate
the yield on and the quantity of its liabilities to determine and alter the price
level.4 This way, the central bank defines the unit of account and determines
all nominal variables. In the frictionless general equilibrium framework of
Arrow and Debreu no-arbitrage conditions ensure that other yields adjust
when the central bank changes the yield on or the quantity of its liabilities.
As the central bank can indirectly manipulate market yields (e.g. the Fed-
eral funds rate or the EONIA rate) to determine the price level, it is possible
in this framework to summarize banks and the central bank by a simple
interest rate rule.3 Demand for central bank liabilities ensures the existence of a general equilibrium in a
monetary economy, i.e., an equilibrium with a positive value for money, see Hahn (1984,Chapters 7 and 8) for details. Gurley and Shaw (1960) described the nature of the demandfor money, but argued that the central bank can also create demand for its liabilities by im-posing minimum reserve requirements on banks. Woodford (2000) argued that the liabilitiesof the central bank define the unit of account (euros, dollars, etc.) for all other contracts thatpeople exchange with each other. This allows central banks to determine the price level evenwithout any demand for its liabilities.
4 Specifically, Patinkin (1961) argued that the central bank can determine the price level byexogenously fixing same nominal quantity (e.g. central bank-liabilities) and some rate ofreturn (e.g. the yield on central-bank liabilities). Accordingly, the central bank can affect andmanipulate the spread between the yield on its liabilities and other market yields to bringabout desired changes in the price level, see also Woodford (2000).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 13PDF page: 13PDF page: 13PDF page: 13
Introduction 3
Apart from theoretical discussions, empirical observations in the early
days of DSGE models seemed to confirm that fluctuations in credit supply
were not a major concern for macroeconomic activity. The insights offered
by Thornton, Wicksell, Fisher and Keynes included a central role for credit
supply because, in their time, credit supply fluctuations had a noticeable
impact on macroeconomic activity. However, the period that directly pre-
cedes the recent global financial crisis was characterized by low inflation
and stable economic growth—the great moderation. In this period, it seemed
that fluctuations in credit supply had almost no impact on economic activ-
ity (Chari et al., 2007). As the role of credit supply was pushed to the back-
ground, the role of other more evident business cycle propagators such as
real and nominal rigidities, the formation of expectations and supply side
shocks dominated the macroeconomic debate.5 The lack of empirical evid-
ence corroborating the importance of fluctuations in credit supply in com-
bination with a theoretical framework in which credit supply is largely ir-
relevant, typically, trivialized the role of credit in macromodels.
The global financial crisis exemplified that deteriorating credit market
conditions are not only a reflection of a depressed real economy, but can also
be a determinant of declining economic activity. In response to the decline
in economic activity central bankers committed to unprecedented levels of
monetary accommodation. However, they could not prevent the global fin-
ancial crisis having a negative impact on the real economy that lasted for
more than a decade. The deterioration of lending conditions in credit mar-
kets was not merely a reflection of a decline in economic activity. In contrast,
the collapse of Lehman Brothers deteriorated lending conditions in credit
markets which caused a decline in credit supply. As mainstream DSGE mod-
els abstracted from credit markets altogether, they were of little use in guid-
ing policy makers on how to act. For this reason, and others, the models
were heavily criticized in the aftermath of the global financial crisis (see e.g.
Caballero (2010), Woodford (2010), Stiglitz (2011) and Romer (2016)).
5 See Mankiw (2006) for a discussion of the development of the economic debate, Lucas(1976) for a discussion of the role of expectation formation and Kydland and Prescott (1982)for the role of supply side shocks in determining macroeconomic fluctuations.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 14PDF page: 14PDF page: 14PDF page: 14
4 Chapter 1
After the outburst of the global financial crisis, the literature started to
take a closer look at the assumptions underlying the framework of Arrow-
Debreu, Modigliani-Miller and Gurley-Shaw. Progress was relatively strong,
because macroeconomists did not have to reinvent the wheel. Thornton,
Wicksell, Fisher, Keynes and their successors already provided a founda-
tion for the role of credit supply in causing macroeconomic fluctuations.
Also the literature on imperfect information with seminal contributions by
Akerlof (1970), Rothschild and Stiglitz (1976) and Townsend (1979) offered
guidance on how to relax the assumptions underlying the framework of
Arrow-Debreu, Modigliani-Miller and Gurley-Shaw. The widespread integ-
ration of frictions in the market for credit in general equilibrium macroe-
conomic models led to a so called second generation of DSGE models with
early contributions by Bernanke et al. (1999) and Kiyotaki and Moore (1997).
1.2 Aim of this thesis
This thesis aims to explore how the availability and allocation of credit af-
fects macroeconomic fluctuations. Throughout this thesis we focus on two
related issues. First, we examine the determinants of credit supply. Credit
supply is essential for economic activity. The recent global financial crisis
led to a global economic crisis when credit markets, most noticeably banks,
stopped lending to households and firms. For the same reason, the euro
area sovereign debt crisis that followed was amplified by financial markets
being unwilling or unable to lend to sovereigns. However, the build-up of
financial imbalances started with credit being structurally mis-allocated. It
is therefore important to assess whether credit is efficiently allocated, espe-
cially when its supply appears unconstrained. We therefore also examine
factors that affect the allocation of credit.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 15PDF page: 15PDF page: 15PDF page: 15
Introduction 5
1.3 Outline
In line with Gurley and Shaw (1960), much of the literature concerning mon-
etary policy is based on the assumption that the central bank directly con-
trols lending conditions in credit markets. The recent global financial crisis
demonstrated, however, that this assumption is not always appropriate. In
particular, when the banking sector is undercapitalized, monetary transmis-
sion can be impeded. To preserve monetary transmission regulators can
implement dynamic loan loss provisioning, a measure taken by the Bank
of England and Banco d’Espana.6 Dynamic provisioning for future credit
losses is forward-looking since banks estimate their expected credit losses
over the business cycle and build up provisions during upswings and draw
down on them during downturns. In contrast, backward-looking provision-
ing relates provisioning to the occurrence of problem loans which has as
potential drawback that expected credit losses are under-provisioned when
the downturn sets in.
Chapter 2 decouples the direct connection between the policy rate and
the real economy to analyze whether dynamic loan loss provisioning could
attenuate macroeconomic fluctuations. To do so, it introduces a banking sec-
tor that is exposed to credit default risk. The representative bank maxim-
izes its expected profits while it anticipates that a fraction of the credit it
supplies will not be paid back in the future. For these expected losses the
banking sector builds up provisions. In order to assess whether loan loss
provisioning affects credit supply, we estimate a panel-VAR for an unbal-
anced sample of 12 OECD countries. The results suggest that unexpected
changes in credit and loan loss provisioning are important drivers of busi-
ness cycle fluctuations. Importantly, loan loss provisioning decreases in re-
lative terms as credit supply increases which suggests that loan loss provi-
sioning is backward looking and therefore amplifies business cycle fluctu-
6 If expected losses are accurately estimated, dynamic provisioning can attenuate fluctu-ations in bank capital and credit supply. In Spain, however, credit supply was (arguably)structurally mis-allocated. Consequently, dynamic provisioning on itself was not sufficientto prevent a severe banking crisis.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 16PDF page: 16PDF page: 16PDF page: 16
6 Chapter 1
ations. In line with the Bank of England and Banco d’Espana, we therefore
advocate forward looking—dynamic—loan loss provisioning.
Whereas Chapter 2 aims to understand how credit default risk impacts
economic activity via loan loss provisioning, Chapter 3 extends this ana-
lysis by also including credit default losses. In the model an unanticipated
increase in credit default losses—a credit default shock—reduces bank cap-
ital. As credit default losses are sunk costs they should not impact the banks’
ability to issue new debt and equity to finance new credit. We fit the model
to euro area data and show, however, that credit default shocks appear to
have been a major driver of historical fluctuations in output via investment.
These results suggest that banks become constrained by their leverage ratio
after a credit default shock and reduce credit supply. Monetary transmission
is impeded because the lending rate increases even though the central bank
lowers the policy rate.
Chapter 3 studies two ex-post measures that could attenuate macroeco-
nomic fluctuations and restore monetary transmission once banks are un-
dercapitalized. Sufficient provisioning or holding large capital buffers for
expected credit default losses (as studied in Chapter 2) could, ex-ante, min-
imize the impact on economic activity as it increases the resilience of the
banking sector. Chapter 3 shows that countercyclical capital buffers as pre-
scribed by the Basel committee on Bank Supervision, i.e., lowering bank
capital regulatory requirements during a downturn and vice versa, can also
be effective in mitigating the effects of a credit default shock ex-post because
it limits the contraction in credit supply. However, there is a trade-off when
the countercyclical capital buffer is activated because banks rebuild their
capital more slowly as leverage constraints bind less. In contrast, a bank
recapitalization financed by lump-sum taxation effectively overcomes this
trade-off problem as rebuilding bank capital is no longer the bank’s choice.
To overcome moral hazard issues associated with bank recapitalizations, we
propose in Chapter 3 to follow the example set by the U.S. government and
recapitalize banks via a bail-in or an equity issuance.
Whereas Chapters 2 and 3 study how monetary transmission can be re-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 17PDF page: 17PDF page: 17PDF page: 17
Introduction 7
stored during a financial crisis to prevent a decline in credit supply, Chapters
4 and 5 focus on the build-up of imbalances that could threaten financial sta-
bility. Especially during more tranquil economic periods, imbalances could
build-up when credit is structurally mis-allocated. When these imbalances
unwind, credit default losses increase, bank capital deteriorates and credit
supply falls which might jeopardize the stability of the financial and eco-
nomic system. Chapters 4 and 5 therefore focus on factors that determine
the allocation of credit.
Chapter 4 studies how an exogenous increase in bank lending is dis-
tributed over mortgage lending versus corporate lending and how this af-
fects growth of the unregulated (shadow) banking sector versus the regu-
lated (traditional) banking sector. Both these developments may adversely
impact future economic activity. On the one hand, a reallocation of credit
supply towards mortgages to finance houses rather than corporate loans
to finance physical capital might harm the economy’s production capacity
making it eventually harder to repay mortgage debt. On the other hand,
a vastly growing shadow banking sector might increase the likelihood of
fire-sales and bank runs. Chapter 4 builds a theoretical model to show how
these two developments might be related. In particular, an exogenous in-
flow of funds in domestic banks depresses real interest rates economy-wide
and increases the share of mortgages in total bank lending. Subsequently,
the interbank market for mortgage loans becomes more liquid which in-
creases shadow banks’ competitive advantage over traditional banks and
gives rise to growth of the shadow banking sector.
Chapter 4 shows that the creation of uninsured shadow bank deposits
increases the banking sector’s reliance on liquidity insurance provided by
the central bank. The banking sector does not fully incorporate the costs of
shadow banks liquidating their assets to accommodate a run because indir-
ectly shadow banks are also insured by the liquidity provision of the central
bank. As the market (partially) ignores liquidity risk, shadow banks create
too many uninsured deposits. In Chapter 4 we argue that one way to realign
private and social interests is by offering households an interest rate on cent-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 18PDF page: 18PDF page: 18PDF page: 18
8 Chapter 1
ral bank money. If this interest rate is set correctly, the incentive for shadow
banks to finance their assets with runnable deposits rather than equity is
removed.
Chapter 5 presents a theoretical framework for a monetary union in
which an exogenous increase in credit supply is typically allocated towards
the less productive nontradable sector (which includes the housing sector).
A fall in the interest rate induces a regional demand boom and increases
demand for both tradable and nontradable goods. Whereas the nontrad-
able sector is able to increase prices and output, the tradable sector faces
foreign competition and thus has less room to increase prices. Therefore, in
real terms, capital and labor are cheaper in the nontradable sector and are
(re)allocated to this sector. As a result, the discrepancy between the external
debt level and the capacity to repay increases. This decreases the solvency
of the recipient region.
We confirm the predictions of the theoretical model by an empirical ana-
lysis that focuses on the euro area. In the decade following the introduction
of the euro, many Southern EMU members experienced a significant fall in
real interest rates and sizeable capital inflows. In a reduced-form Bayesian
panel-VAR for 10 euro area countries, the countries which experienced neg-
ative interest rate shocks (relative to the euro area average) are shown to
experience faster growth of the nontradable sector which contributed to a
deteriorating current account balance. The same reduction in interest rates,
however, did not affect growth of the tradable sector.
The allocation of credit in Chapters 4 and 5 has one important element
in common: in both chapters credit for the housing sector expands. While in
Chapter 4 the supply elasticity of the underlying asset that serves as collat-
eral determines the allocation of credit, in Chapter 5 the absence of foreign
competition for the nontradable sector allows this sector to expand. Thus,
sectors that produce goods that are both nontradable and have a low sup-
ply elasticity are therefore expected to expand and to show large price in-
creases when credit supply increases. Clearly the housing sector is the most
prevalent sector that falls in both categories.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 19PDF page: 19PDF page: 19PDF page: 19
Introduction 9
These results raise an important policy issue as to prevent excessive bor-
rowing for residential housing. A vastly growing housing sector financed
with mortgage loans could have adverse effects on future economic growth
because the economy’s production capacity does not grow in accordance
with its debt level and the housing sector is typically associated with low
productivity growth. For these reasons, some countries have introduced
constraints on admissible loan-to-value (LTV) ratios to create a precaution-
ary buffer against a decline in house prices. Chapter 4 shows that tighter
LTV constraints can indeed attenuate house price and mortgage supply fluc-
tuations. Chapter 5 suggests to liberalize the tradable sector by increasing
competition in this sector. Both interventions, tighter LTV constraints for
mortgage loans and a liberalization of the tradeable sector, reallocate funds
from the nontradeable, housing sector to more productive sectors and might
thereby benefit future economic growth and financial stability.
Finally, Chapter 6 summarizes and concludes.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 20PDF page: 20PDF page: 20PDF page: 20
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 21PDF page: 21PDF page: 21PDF page: 21
Chapter 2
Loan loss provisioning, bank
credit and the real economy∗
2.1 Introduction
The recent global financial crisis has shown that bank credit is an important
determinant of business cycle fluctuations. Before the crisis, bank credit was
abundant (Adrian and Shin, 2009), boosting economic growth. During the
crisis, credit default risk, i.e., the risk that a borrower is unable to pay back
a bank loan, increased, restraining the issuance of new bank loans.
In this chapter we examine how credit default risk affects bank lending
and the business cycle. As a measure of credit default risk, we use loan loss
provisioning by banks. While most of the literature on loan loss provision-
ing examines its determinants, we are especially interested in how loan loss
provisioning affects credit and the real economy. Until now the effect of loan
loss provisioning on the real economy only received limited attention in the
literature. Furthermore, most previous studies use bank-level data instead
of macro data.
Bank loan loss provisioning may be either procyclical or countercyclical,
depending on whether provisioning is backward-looking (sometimes called
‘non-discretionary’) or forward-looking (‘discretionary’). Backward-looking
∗This chapter is based upon Pool et al. (2015).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 22PDF page: 22PDF page: 22PDF page: 22
12 Chapter 2
provisioning relates provisioning to the occurrence of problem loans. This
has as potential drawback that expected credit losses are underprovisioned
during upswings, when few problem loans are identified and hence the
level of provisioning is low. Conversely, during downturns provisioning
increases because credit defaults are high. As a result, backward-looking
provisioning is procyclical.1 In contrast, forward-looking provisioning is
countercyclical. Banks estimate their expected credit losses over the busi-
ness cycle and build up provisions during upswings and draw down on
them during downturns.
Accounting rules contribute to backward looking provisioning, as they
tend to allow provisions based on past events, not on expectations (Borio
and Lowe, 2001). International Financial Reporting Standards (IFRS) utilize
a so-called incurred loss model where loan losses are recognized only after
loss events have occurred prior to the reporting date that are likely to result
in future non-payment of loans. This is the so-called International Account-
ing Standard (IAS) 39 rule under IFRS. This rule does not allow for consid-
eration of future expected losses based on trends suggestive of additional
future losses (Bushman and Williams, 2012).
Most of the empirical finance literature confirms backward-looking pro-
visioning. Bikker and Metzemakers (2005) find evidence of a negative rela-
tion between GDP growth and provisioning for 29 OECD countries, imply-
ing backward looking practices. This procyclicality is mitigated partly by
the positive relation between banks earnings and provisions, which might
be due to either income smoothing or forward looking provisioning. Laeven
and Majnoni (2003) also find evidence that banks around the world are less
prudent during periods of rapid credit growth, in the sense that under fa-
vorable conditions banks postpone provisioning until unfavorable condi-
tions set in. Bouvatier and Lepetit (2008) examine the impact of loan loss
provisions on bank lending using a sample of 186 European banks for the
period 1992-2004. They find that backward looking provisioning amplifies
1 Bolt et al. (2012), using aggregate bank data for an unbalanced set of 17 countries over theperiod 19792007, find that loan losses are the main driver of the negative impact of recessionson bank profits.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 23PDF page: 23PDF page: 23PDF page: 23
Loan loss provisioning, bank credit and the real economy 13
credit fluctuations, while forward looking provisioning or income smooth-
ing does not. Empirical work by Jimenez et al. (2012), examining the impact
of countercyclical capital buffers on credit supply using countercyclical dy-
namic provisioning experiments in Spain, find that countercyclical capital
buffers help smooth credit supply cycles.
The usefulness of loan loss provisioning for macroprudential regulation
has recently also received attention in the theoretical literature. Bouvatier
and Lepetit (2012), in a partial equilibrium framework, show that forward-
looking provisions can eliminate procyclicality in lending standards induced
by backward-looking provisions. Agenor and Zilberman (2015), in a calib-
rated DSGE model, show that forward-looking loan loss provisions can re-
duce volatility in financial and real variables by mitigating the changes in
the stock of loan-loss reserves over the course of the business cycle. Zil-
berman and Tayler (2014) examine the interaction between loan loss provi-
sioning rules, business cycle fluctuations and monetary policy in a DSGE
model with endogenous credit risk. These authors highlight the importance
of forward-looking provisions in mitigating welfare losses, as well as how
accounting rules with respect to loan loss provisions alter the transmission
mechanism of monetary policy.
For our analysis we set up a macroeconomic framework including a
banking sector and credit default risk. The aim of the model is to under-
pin our empirical panel-VAR model with a theoretical framework. We sim-
plify an established theoretical framework to bring the model to the data.
To keep the number of variables tractable, we use an industrial organization
approach to model the banking sector (Freixas and Rochet, 1997). The rep-
resentative bank maximizes its expected profits anticipating that a fraction
of credit will default in the future (Greenbaum et al., 1989). We implicitly
solve for the optimal levels of credit and the lending rate (i.e., the price of
credit), given the short-term interest rate (the cost of credit) and credit de-
fault risk. The equilibrium conditions for credit and the lending rate are em-
bedded in a standard closed-economy macroeconomic framework, as often
used to analyze monetary transmission (see e.g., Svensson 1997 and Clarida
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 24PDF page: 24PDF page: 24PDF page: 24
14 Chapter 2
et al. 1999). Hence, instead of assuming a perfect interest rate pass-through,
credit risk and market power in the banking sector determine the interest
rate spread (in line with Christiano et al. (2014) for the former and Berger
et al. (2004) for the latter). The representative bank is exposed to credit risk
which imposes a potential cost for the bank. Consequently, an increase in
credit default risk increases the lending rate and decreases bank lending.
In order to assess whether the data support our theoretical model, we
estimate a panel-VAR for an unbalanced sample of 12 OECD countries over
the last two or three decades (1980/1990-2008/9); the sample is determined
by the availability of macroeconomic provisioning data.2 Panel VARs can
be used to uncover the dynamic relationships that are common to all cross-
sectional units.3
Our panel-VAR impulse response functions (IRFs) are generally in line
with our theoretical model. First, the results suggest that credit risk (meas-
ured by provisioning by the banking sector) is one of the drivers of busi-
ness cycle fluctuations. Specifically, an increase in provisioning decreases
bank lending and economic activity. Second, it appears that banks decrease
provisioning as a percentage of total bank assets when bank lending in-
creases and vice versa. Hence, during upswings banks take on more risk by
building up relatively low provisions while in downswings, banks build up
loan loss provisions. These results confirm backward-looking provisioning.
Third, output is an important determinant of bank lending, more so than
other factors such as interest rates.
The remainder of the Chapter is structured as follows. Section 2.2 de-
scribes the theoretical model, Section 2.3 the data and Section 2.4 presents
the results. Section 2.5 concludes.2 After 2009 the OECD discontinued the publication of the Bank Profitability Statistics from
which the macroeconomic provisioning data are taken.3 For example, Love and Zicchino (2006) study the impact of financial factors on firm in-
vestment and de Haan and van den End (2013) examine banks responses to market fundingshocks. See Canova and Ciccarelli (2013) for a survey of the panel-VAR literature.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 25PDF page: 25PDF page: 25PDF page: 25
Loan loss provisioning, bank credit and the real economy 15
2.2 The model
This section presents the macroeconomic framework, discusses the model
predictions of a loan loss provisioning shock and describes the empirical
set-up.
2.2.1 Macroeconomic framework
Our framework is closely related to DSGE models that examine the trans-
mission of credit risk to business cycles, see e.g., Bernanke et al. (1999), Cur-
dia and Woodford (2010) and Christiano et al. (2014). As mentioned in the
Section 2.1, the aim of the model is to underpin the empirical panel-VAR
model with a theoretical framework.
The banking sector acts as an intermediary sector which lends to the real
economy at lending rate ilt and funds itself by short-term debt with a (risk-
free) short-term interest rate ist ; see Freixas and Rochet (2008). The difference
between the short-term interest rate and the lending rate consists of the term
spread and a credit risk premium. We follow the conventional approach by
assuming that the term spread is constant over time (e.g., Woodford and
Walsh 2005). Bouvatier and Lepetit (2012) show theoretically that credit risk
(accounted for via backward-looking loan loss provisioning) affects the loan
rate through at least two specific channels. If the realized number of credit
defaults is higher than anticipated, the loan rate increases because (i) expec-
ted future interest earnings decrease and (ii) the unanticipated loss deterior-
ates banks’ capital. The first effect works directly through the risk premium.
Banks update their beliefs about future defaults and increase loan loss provi-
sioning. The second effect works via the banks balance sheet. Banks increase
their loan loss provisioning to cover unanticipated losses. The decrease in
the banks capital position increases the lending rate. Here we focus on the
credit risk premium channel and leave the balance sheet channel to future
research.
Imagine a representative bank which maximizes profits from lending
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 26PDF page: 26PDF page: 26PDF page: 26
16 Chapter 2
activities:
jt =(
iltϑt − is
t
)cn
t , (2.1)
where jt represents bank profits, ϑt the expected credit repayment rate, and
cnt denotes new credit and subscripts denote the time index.4 The difference
between the short-term interest rate ist (the cost price for the bank) and the
risk adjusted lending rate iltϑt is determined by the credit spread.
The credit demand curve is constructed as follows. First, we assume that
demand for new credit cnt depends negatively on the price of credit, i.e.,
the lending rate. Second, demand for new credit depends positively on the
business cycle as measured by the output gap yt. Third, demand for new
credit depends positively on the price level pt, since a high inflation rate
reduces the real interest rate ceteris paribus.5 Taking the inverse of this (by
assumption invertible) relationship with respect to the lending rate ilt, we
obtain the following relation denoted by f (·):
ilt = f
(cn
t−k, yt−k, pt−k)
, k = 0, 1, ..., q, (2.2)
where q denotes the number of lags we consider. Substituting the expres-
sion for the lending rate (2.2) into (2.1) and maximizing with respect to new
credit, yields (see Appendix 2.A):
f(cn
t−k, yt−k, pt−k)=
1ϑt
(ist − f
′ (cn
t−k, yt−k, pt−k)
cnt
), (2.3)
where f′cn denotes the derivative of f (·) with respect to cn
t . Equation (2.3)
gives the relation between the short-term interest rate and the real economy.
It follows from (2.1) and (2.2) that the lending rate depends positively on the
short-term interest rate, ∂ilt
∂ist> 0, negatively on the expected credit repayment
4 We assume financing costs to be independent from the risk level of the banks existing bal-ance sheet. Our representative bank can always finance itself by borrowing from the centralbank at rate is
t .5 Demand for new credit may also depend on other variables not endogenously determined
in the model. We do not consider exogenous variables in our framework.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 27PDF page: 27PDF page: 27PDF page: 27
Loan loss provisioning, bank credit and the real economy 17
rate, ∂ilt
∂ϑt< 0, and negatively on the amount of new credit, ∂il
t∂cn
t< 0.
For ϑt, the expected credit repayment rate (2.1), we assume that banks
expect the payback rate in the next period to be equal to the payback rate
observed in the current period. This assumption of static expectations form-
ation is consistent with the empirical evidence of backward looking pro-
visioning behavior found by Laeven and Majnoni (2003) and Bikker and
Metzemakers (2005). The expected credit repayment rate is therefore:
ϑt =ct − Etdt+1
ct=
ct − dt
ct, (2.4)
where ct denotes total credit, dt denotes credit defaults, and Et· is the
expectation operator. Substituting (2.4) into (2.3) yields:
f(cn
t−k, yt−k, pt−k)=
ct
ct − dt
(ist − f
′cn(cn
t−k, yt−k, pt−k)
cnt
). (2.5)
Since (2.5) is an implicit function for new credit, we can only implicitly solve
it for new credit and substitute the solution into the lending rate function,
(2.2). Hence, we replace cnt in (2.2) by the solution of (2.5) and find that the
lending rate is defined as a function g(·) of the following variables, see Ap-
pendix 2.A:
ilt = g(dt−k, yt−k, pt−k, is
t−k, ct−k). (2.6)
We assume that the law of motion for credit, ct, equals new credit minus
credit defaults plus the share of credit that does not mature, λ:
ct = λct−1 − dt + cnt , 0 < λ < 1, (2.7)
where the credit shock εct is incorporated in cn
t ; see Equation (2.A.7) in Ap-
pendix 2.A.6 We assume that the credit default variable, dt, follows a sta-
tionary AR(1) process that returns to its equilibrium value, a percentage δ of
6 Our representation of the banking sector is short-term oriented. We do not take into ac-count, for example, that prudential provisioning might increase credit in the long-term.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 28PDF page: 28PDF page: 28PDF page: 28
18 Chapter 2
total credit:
dt = (1− ρ)δct−1 + ρdt−1 + εpt , 0 < ρ < 1. (2.8)
Equation (2.8) states that credit defaults are a fraction of total credit in the
previous period, (1− ρ)δ, and a fraction ρ of the amount of credit defaults
in the previous period. Hence, ρ captures the persistence of credit defaults,
and (1− ρ) determines how fast the number of credit defaults returns to the
average default rate δ after a shock, denoted by εpt . The shock ε
pt is labeled a
provisioning shock, since we use bad loan provisioning data to proxy credit
default risk.
Using the solution of (2.5-2.8) we can solve the equation for total credit,
see Appendix 2.A. The model tries to estimate the effects of credit risk on
economic activity. The risk premium is linked to the degree of credit risk
and the risk-free part of the lending rate is captured by the short-term in-
terest rate. The term premium is kept constant, as mentioned above. We log-
linearize the lending rate function (2.6), the equation for total credit (2.7)
and credit defaults (2.8). Throughout this chapter, variable symbols with a
hat represent log-linearized variables, except for the interest rates ilt and is
t .
For ilt and is
t , we use that log-linearization of an interest rate under the as-
sumption that its steady state equals zero (i = 0), yields: (1+it)−(1+i)1+i = it,
which we denote by it.
The log-linearized solutions to the lending rate function, total credit and
credit defaults are embedded in a standard closed economy macroeconomic
framework, complemented by generalized versions of the aggregate de-
mand curve (2.9), the Philips curve (2.10) and the Taylor rule (2.11):
yt =Φ1(L)yt + Φ2(L)(ilt − πt) + εa
t , (2.9)
πt =Φ3(L)πt + Φ4(L)yt + εst, (2.10)
ist =γyt + ϕπt + εm
t , (2.11)
where πt denotes the inflation rate, and Φj(L) is a lag polynomial Φj(L) ≡
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 29PDF page: 29PDF page: 29PDF page: 29
Loan loss provisioning, bank credit and the real economy 19
Φj,1L1 + ... + Φj,qLq for j = 1, 3 and Φj(L) ≡ Φj,0 + Φj,1L1 + ... + Φj,qLq for
j = 2, 4 where q denotes the number of lags we consider. The shocks, εat , εs
t
and εmt are labeled Aggregate Demand (AD) shock, Cost Push (CP) shock,
and Monetary Policy (MP) shock, respectively. The aggregate demand curve
(2.9) describes the relationship between the output gap and the real lending
rate, ilt − πt, the Philips curve (2.10) the relationship between the inflation
rate and the output gap, and the Taylor rule (2.11) the relationship between
the short-term interest rate and the inflation rate and the output gap.
Using (2.6) to substitute out the lending rate variable and imposing re-
strictions on the contemporaneousness of shocks and responses (see below),
we summarize the model as a structural Vector Auto Regressive (VAR) sys-
tem:
A(L)Zt = εt (2.12)
where we assume that εt is iid ∼ (0, Σε), Σε = Eεt, ε′t, and A(L) is a
lag polynominal of the form A(L) = A0 − A1L − ...− ApLp, in which Ak,
k = 1, ..., p, are coefficient matrices. We rewrite (2.12) into a reduced form:
Zt = B1Zt−1 + B2Zt−2 + ... + BpZ−p + vt (2.13)
where Bk ≡ A−10 Ak, vt ≡ A−1
0 εt, and vt is iid ∼ (0, Σv), Σv = Evt, v′t. We
define the vectors as follows:
Zt =[dt yt πt is
t ct
]′and εt =
[ε
pt εa
t εst εm
t εct]′
, (2.14)
where we use that: pt−pp = πt which we denote as πt.
As the reduced form disturbances, vt, represent the effect of all struc-
tural shocks in the economy, it is not possible to ascribe a particular struc-
tural shock in εt, for example a MP shock, to vt (Christiano et al., 1999).
Therefore, for identification of the structural shocks it is common practice
to assume, first, that the structural shocks are orthogonal, i.e., Σε is a diag-
onal matrix with the standard deviations on the diagonal. Second, one has
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 30PDF page: 30PDF page: 30PDF page: 30
20 Chapter 2
to make identification assumptions to identify the relationship between the
reduced form VAR disturbances, vt , and the structural shocks, εt.
We use (2.13) to identify this relationship, i.e., εt = A0vt ∼ (0, Σε =
A0ΣvA′0), where A0 is the invertible square matrix in (2.12). Since A0 is a
lower triangular matrix, the structural shocks in (2.13) are identified by as-
suming a recursive system which imposes zero restrictions on all elements
of A0 above the diagonal, which is also known as a Cholesky Decomposi-
tion.
In particular, our model assumes a number of restrictions with respect
to the contemporaneous shocks and responses. The output gap, yt, is only
contemporaneously affected by provisioning and AD shocks. There is in-
deed considerable consensus in the literature that the output gap is only
modestly affected by shocks in other variables (e.g., Bernanke and Gertler
(1995) and Christiano et al. (1999)).
Inflation, πt, is assumed to be only contemporaneously affected by pro-
visioning, AD and CP shocks. The literature often assumes that prices re-
spond very sluggishly to shocks in other variables (for example, Bernanke
and Gertler (1995) and Christiano et al. (1999)).
The short-term interest rate, ist , is assumed to be contemporaneously af-
fected by provisioning, AD, CP and MP shocks.
Credit, ct, is contemporaneously affected by provisioning, AD, CP, MP
and credit shocks since new credit, cnt , contains the contemporaneous vari-
able ist . Banks assess the most recent data available to determine credit.
Credit defaults, dt, are only contemporaneously affected by provision-
ing shocks; other shocks have an impact after one period. This assumption
reflects backward-looking provisioning behavior, as empirically confirmed
by e.g., Laeven and Majnoni (2003) and Bikker and Metzemakers (2005).
2.2.2 What does a provisioning shock do?
This section discusses the predicted effects of an unanticipated change in
loan loss provisioning. The model presented in this section contains reduced
form equations and implicit functional forms. We cannot use structural para-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 31PDF page: 31PDF page: 31PDF page: 31
Loan loss provisioning, bank credit and the real economy 21
meters to calculate reduced form coefficients and generate theoretical im-
pulse response functions. Instead we describe the comparative statics of the
credit market after a provisioning shock in Figure 2.1.
Loan loss provisioning increases because banks expect a lower repay-
ment rate. Therefore, banks will try to compensate the expected loss by de-
creasing credit supply, see (2.6). As a consequence, credit decreases after a
positive provisioning shock, see (2.7). This is represented by a movement of
the credit supply curve in Figure 2.1 from cs0 to cs
1. The increase in the lending
rate decreases the output gap via the aggregate demand curve which causes
the inflation rate to decrease via the Philips curve. The drop in the output
gap and the inflation rate causes the credit demand curve in Figure 2.1 to
shift from cd0 to cd
1. As a consequence, the economy moves from (c0; ll0) to
(c1; ll1). Note that the total amount of credit in the economy falls unambigu-
ously, whereas the lending rate can either increase or decrease depending
on the elasticities of the credit demand and credit supply curve.
2.2.3 Empirical setup
To bring the model to the data, we estimate a reduced form panel-VAR sys-
tem adding country specific fixed effects:
Zi,t = ui + B(L)Zi,t + vi,t, (2.15)
where Zi,t is a vector of endogenous variables, i = 1, 2, ..., 12 denotes the
country index, ui is a vector of country-specific fixed effects, B(L) is a lag
polynomial B(L) ≡ B1L1 + ... + BpLp, and vi,t is a vector of stacked reduced
form residuals. The vector Zi,t consists of the endogenous variables intro-
duced in Section 2.2.1 stacked per country, Zi,t =[di,t, yi,t, πi,t, is
i,t, ci,t
]′.
The main advantage of using a panel approach is the increased efficiency
of statistical inference. High-frequency macroeconomic provisioning data
are not available and thus the number of observations is relatively small. In
VAR models the number of coefficients increases with the number of vari-
ables squared. Estimating a 5-variable VAR lacks degrees-of-freedom if time
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 32PDF page: 32PDF page: 32PDF page: 32
22 Chapter 2
Figure 2.1. Static representation of an increase in provisioning.
il
c
cs0
cs1
cd0cd
1
il0
c0
il1
c1Note: the lending rate is depicted on the vertical axis and credit is depicted on the horizontalaxis. cd
τ and csτ denote the credit demand and credit supply curve at point τ = 0, 1 in time,
respectively.
series have low frequency. To overcome the degrees-of-freedom issue, we
use a panel-VAR approach. The panel-VAR approach implicitly imposes the
same underlying structure to each country in the panel. Cross-country het-
erogeneity is allowed for by adding individual fixed effects. As mentioned
in the Section 2.1, our model is macro-oriented and our focus is not on the
determinants of loan loss provisioning or income smoothing, but on the ef-
fect of loan loss provisioning on the macro-economy.7
7 For example, we assume that institutional differences between countries are time-invariant. Other, micro-oriented studies focus more on the determinants of loan loss pro-visioning. For example, Fonseca and Gonzalez (2008), using micro-data for 3221 bank-yearobservations from 40 countries, present evidence that income smoothing by managing loan
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 33PDF page: 33PDF page: 33PDF page: 33
Loan loss provisioning, bank credit and the real economy 23
We estimate System (2.15) using the Generalized Method of Moments
(GMM). The fixed effects are eliminated by expressing all variables as de-
viation from their means. Since the fixed effects are correlated with the re-
gressors as a result of the inclusion of lags of the dependent variables, or-
dinary mean-differencing (i.e., expressing all variables as deviations from
their full sample periods means) as commonly used to eliminate fixed ef-
fects would create biased coefficients. To avoid this problem, forward mean-
differencing, also known as Helmert transformation, is used instead (cf.
Arellano and Bover 1995). This procedure removes only the forward mean,
i.e., the mean of all future observations available in the sample and pre-
serves the orthogonality between transformed variables and lagged regres-
sors, so that the lagged regressors can be used as valid instruments for es-
timating the coefficients by system GMM.8
2.3 Data
Our sample includes 12 OECD countries: Austria, Belgium, Denmark, Fin-
land, France, Germany, Italy, Japan, the Netherlands, Spain, Sweden, and
the United States. We selected developed western economies with a relat-
ively high degree of homogeneity, for which data availability, notably with
respect to loan loss provisions, was no problem. We use annual time series
of the OECD for the output gap, inflation, the short-term interest rate, out-
standing bank loans to the private sector and loan loss provisions. For de-
tails, see Table 2.B.1 in Appendix 2.B.
Expectations with respect to credit defaults is a latent variable, which
we proxy by banks’ loan loss provisioning. The provisions data series starts,
depending on the country, between 1979 and 1988 and ends either in 2008
or 2009. In order to make country comparison feasible we transform this
variable by taking the percentage of loan loss provisioning to the total bank
loss provisions depends on investor protection, disclosure, regulation and supervision, fin-ancial structure, and financial development.
8 For more details about the estimation procure we refer to Love and Zicchino (2006), whoseStata code we gratefully use for our estimation.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 34PDF page: 34PDF page: 34PDF page: 34
24 Chapter 2
Table 2.1. Summary statistics for loan loss provisions, per country.
Countries Years Mean Standard deviation Median
Austria 1989-2008 0.031 0.223 -0.024Belgium 1981-2009 -0.008 0.133 0.000Denmark 1979-2009 -0.003 0.322 -0.001Finland 1979-2009 -0.012 0.122 -0.005France 1988-2009 0.000 0.122 0.017Germany 1979-2009 0.007 0.113 -0.001Italy 1984-2009 -0.008 0.127 -0.026Japan 1989-2008 0.009 0.299 0.003Netherlands 1979-2009 -0.012 0.129 -0.009Spain 1979-2009 0.012 0.212 0.002Sweden 1979-2009 -0.028 0.912 0.033United States 1980-2009 0.057 0.252 0.009
Note: First difference of loan loss provisions as percentage of total bank assets.
balance sheet. Table 2.1 reports the descriptive statistics of loan loss provi-
sioning as percentage of total bank assets. The provisions series of France
does not start before 1988, while those of several other countries start in
1979. Provisioning is only a small percentage of the total balance sheet. Fig-
ure 2.2 shows that especially during the years before the global financial
crisis of 2008, provisioning levels were historically low for most countries
while during the global financial crisis provisioning levels started to rise
sharply.9 Table 2.2 shows the summary statistics for all transformed vari-
ables. The dimensions of the variables are: first difference of loan loss provi-
sions as percentage of total bank assets, output gap as percentage deviation
of its trend, inflation rate in percentages (first differences of logs of the price
level multiplied by 100%), short-term interest rate in levels, and credit in
9 Figure 2.2 shows that, during the late 1980s and early 1990s, loan loss provisioning inSweden, experiencing a banking crisis during the time, declines sharply. Because of this pe-culiarity, Bolt et al. (2012) drop Sweden from their sample. Results, which are not presentedhere but are available on request, show that our main findings do not change significantlywhen Sweden is omitted from the sample.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 35PDF page: 35PDF page: 35PDF page: 35
Loan loss provisioning, bank credit and the real economy 25
Figure 2.2. Loan loss provisions as a percentage of the total balance sheet ofthe banking sector, annual by country.
-0,5%
0,0%
0,5%
1,0%
1,5%
1979 1984 1989 1994 1999 2004 2009
(A) Austria
-0,2%
0,0%
0,2%
0,4%
0,6%
0,8%
1979 1984 1989 1994 1999 2004 2009
(B) Belgium
0,0%
0,5%
1,0%
1,5%
2,0%
1979 1984 1989 1994 1999 2004 2009
(C) Denmark
-0,2%
0,0%
0,2%
0,4%
0,6%
0,8%
1979 1984 1989 1994 1999 2004 2009
(D) Finland
-0,2%
0,0%
0,2%
0,4%
0,6%
0,8%
1979 1984 1989 1994 1999 2004 2009
(E) France
0,0%
0,2%
0,4%
0,6%
0,8%
1979 1984 1989 1994 1999 2004 2009
(F) Germany
0,0%
0,2%
0,4%
0,6%
0,8%
1979 1984 1989 1994 1999 2004 2009
(G) Italy
0,0%
0,5%
1,0%
1,5%
1979 1984 1989 1994 1999 2004 2009
(H) Japan
0,0%
0,2%
0,4%
0,6%
0,8%
1,0%
1979 1984 1989 1994 1999 2004 2009
(I) Netherlands
0,0%
0,5%
1,0%
1,5%
1979 1984 1989 1994 1999 2004 2009
(J) Spain
-4,0%
-2,0%
0,0%
2,0%
1979 1984 1989 1994 1999 2004 2009
(K) Sweden
0,0%
0,5%
1,0%
1,5%
2,0%
1979 1984 1989 1994 1999 2004 2009
(L) United States
percentage changes (first differences of the logs of total credit multiplied by
100%).
To test whether the series contain unit roots, we performed Levin et al.
(2002) panel data unit root tests after conversion into balanced panels. We
do this for the series suppressing panel-specific means, as our panel-VAR
model assumes fixed country effects so that the relevant variables to look at
are the variables after removing the panel means. The results show that all
series are stationary, see Table 2.3.10
2.4 Results
We present the main results followed by some robustness checks.
10 Alternatively, Im et al. (2003) tests for unbalanced panels confirm stationarity of all modelvariables.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 36PDF page: 36PDF page: 36PDF page: 36
26 Chapter 2
Table 2.2. Summary statistics for the model variables.
Variables Obs. Mean Standard deviation Median
dt 321 0.010 0.332 0.002yt 329 -0.104 2.411 0.029πt 527 -0.004 2.095 0.000ist 552 6.241 4.773 5.278
ct 492 8.645 6.196 8.693Note: First difference of loan loss provisions as percentage of total bank assets; output gapas percentage deviation of its trend; inflation rate in percentages (∆ logs of the price levelmultiplied by 100%); short-term interest rate in levels; credit in percentage change (∆ logs oftotal credit multiplied by 100%).
Table 2.3. Levin et al. (2002) unit-root test.
Variable Adjusted t p-Value
yt -10.97 0.00πt -6.71 0.00ct -4.84 0.00ist -5.68 0.00
dt -10.50 0.00Note:H0: Panels contain unit roots. Ha: Panels are stationary. ADF regression: 4 lags, ARparameter: common. LR variance: Bartlett kernel. Panel means not included.
2.4.1 Main results
The panel-VAR is estimated including 1 lag in line with the Akaike and
Schwarz information criteria for the individual time series. Estimation res-
ults, which for reasons of space are not presented but are available on re-
quest, prove to be robust to different lag length specifications. Instead, as
is the convention for VAR models, impulse-response functions (IRFs) are
presented.
All shocks are labeled as specified in Section 2.2. Following Jacobs and
Wallis (2005) we apply directly interpretable impulse magnitudes instead
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 37PDF page: 37PDF page: 37PDF page: 37
Loan loss provisioning, bank credit and the real economy 27
of the conventional one standard deviation shocks, which depend on the
fit of the equations of the VAR model. The IRFs presented show the first 6
periods after the shock with 90% confidence intervals generated by Monte-
Carlo with 1000 iterations.11
This section discusses the main responses of an output gap shock (here-
after: Aggregate Demand (AD) shock) represented by a 1 percentage point
increase in the output gap, a credit shock represented by a 5 percentage
point increase in the credit growth rate, and a provisioning shock set equal
to an increase in the change in provisioning as percentage of total bank as-
sets by 0.2 percentage points. Figure 2.2 shows that many countries exper-
ienced an increase in provisioning close to 0.2 percentage point during the
beginning of the global financial crisis. In addition, Table 2.1 shows that for
many countries the standard deviation of provisioning to total bank assets
is close to 0.2. For these reasons the provisioning shock is set to a 0.2 per-
centage point increase.
The main consequences of a positive provisioning shock represented in
Figure 2.3 are a decrease of the output gap and credit (see panels B1 and
C1, respectively). The output gap declines for more than three years sug-
gesting that provisioning shocks drive business cycle fluctuations. Specific-
ally, a 0.2 percentage point increase in provisioning decreases the output
gap by approximately 0.25 percentage point suggesting a significant de-
cline in economic activity. The effect on credit becomes insignificant after
the first period. Hence, the effect of a provisioning shock on credit has no
long-lasting effects.12
Provisioning itself appears to decrease slightly three years after an AD
shock, but decreases strongly after a credit shock; see panel A2 and A3, re-
spectively. These results suggest that banks do not use economic outlook
measures to determine loan loss provisioning. The model suggests, by con-
11 We experimented with a larger number of iterations and obtained similar results.12 The IRFs of a panel-VAR excluding the provisioning variable are almost identical for thecore model variables (output gap, inflation, short-term interest rate and credit). Hence, thedestabilizing effect credit risk has on the business cycle comes from credit risk shocks, i.e.,provisioning shocks, itself, and does not affect the dynamic relations between the other vari-ables.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 38PDF page: 38PDF page: 38PDF page: 38
28 Chapter 2
Figure 2.3. Impulse response functions for provisioning, Aggregate Demand(AD) and credit shocks, annual data.
-0.2
0.0
0.2
0.4
0 3 6
(A1) Provisioning to provisioning
shock
-0.04
-0.02
0.00
0.02
0.04
0 3 6
(A2) Provisioning to AD shock
-0.3
-0.2
-0.1
0.0
0.1
0 3 6
(A3) Provisioning to credit shock
-0.6
-0.4
-0.2
0.0
0.2
0 3 6
(B1) Output gap to provisioning
Shock
0.0
0.4
0.8
1.2
0 3 6
(B2) Output gap to AD shock
0.0
1.0
2.0
3.0
0 3 6
(B3) Output gap to credit shock
-1.2
-0.8
-0.4
0.0
0.4
0 3 6
(C1) Credit to provisioning shock
-1.0
0.0
1.0
2.0
0 3 6
(C2) Credit to AD shock
0.0
2.0
4.0
6.0
0 3 6
(C3) Credit to credit shock
Note: 90% confidence intervals generated by 1000 Monte-Carlo iterations; periods in yearson the horizontal axis.
struction, that provisioning increases after a positive credit shock because
banks provision a fixed percentage of credit, δ > 0. Our finding is in line
with the empirical evidence in the literature. Cavallo and Majnoni (2002)
find for non-G10 countries a negative correlation between pre-provisioning
income and provisioning. Laeven and Majnoni (2003) present evidence that
banks delay provisioning in good times. As a consequence, provisioning
levels are too low during bad times.
The main consequence of an AD shock represented in Figure 2.3 is an
increase in credit (panel C2). The AD shock raises credit for more than 6
years. The results are in line with our theoretical framework which predicts
an increase in credit during periods of high economic activity. In addition,
the results suggest that an increase in the output gap has long-lasting ef-
fects on credit. The economic impact of the AD shock on credit is large: a 1
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 39PDF page: 39PDF page: 39PDF page: 39
Loan loss provisioning, bank credit and the real economy 29
percentage point increase in the output gap increases credit growth by 1.5
percentage point on impact.
The main consequence of a positive credit shock, represented by a 5 per-
centage point increase in credit, is a persistent increase in the output gap up
to 1.5 percentage point (panel B3). An increase in credit supply, making more
funds available for the purchase of goods, increases aggregate demand. It
appears that credit is an important determinant of economic activity.
To determine which variables drive credit supply, we also investigate
the consequences of a CP shock (1 percentage point increase in the inflation
rate) and a MP shock (100 basis point increase in the short-term interest
rate) on credit supply. The results presented in Figure 2.4 show that credit
is unaffected by CP and MP shocks (panels C1 and C2, respectively). The
results in Figs. 2.3 and 2.4 suggest that credit is mainly affected by an AD
shock; hence, it appears that credit is primarily demand-driven.
2.4.2 Robustness
In this section we show the robustness of our findings for a different defin-
ition of the output gap and for the frequency of the observations, respect-
ively.13 Output gap measures are controversial because the output gap is
hard to estimate. To check for robustness, we replace the OECD output gap
by an output gap measure that we derived by application of the Hodrick-
Prescott (HP) filter on real annual GDP as a measure of potential output. Fig-
ure 2.C.1 in Appendix 2.C shows the same IRFs using this output gap meas-
ure which can be compared with Figure 2.3. The main results remain intact.
However, while the response of the output gap to a provisioning shock is
stronger, the response of the output gap to a credit shock becomes insigni-
ficant. Specifically, the decline in the output gap after a positive provisioning
shock is four times as large as the corresponding decline in 2.3, but lasts only
two years (compare panels B1).
13 We also tested the robustness for different selections of countries in our sample and forsub-sample periods. Results, which are not presented for reasons of space but are availableon request, show that the main conclusions remain the same.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 40PDF page: 40PDF page: 40PDF page: 40
30 Chapter 2
Figure 2.4. Impulse response functions for Cost Push (CP) and MonetaryPolicy (MP) shocks, annual data.
-0.5
0.0
0.5
1.0
1.5
0 3 6
(A1) Inflation to CP shock
-0.8
-0.4
0.0
0.4
0 3 6
(B1) Output gap to CP shock
-1.0
-0.5
0.0
0.5
1.0
0 3 6
(C1) Credit to CP shock
0.0
0.4
0.8
1.2
0 3 6
(A2) Interest to MP shock
-0.6
-0.4
-0.2
0.0
0.2
0 3 6
(B2) Output gap to MP shock
-2.0
-1.0
0.0
1.0
0 3 6
(C2) Credit to MP shock
Note: 90% confidence intervals generated by 1000 Monte-Carlo iterations; periods in yearson the horizontal axis.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 41PDF page: 41PDF page: 41PDF page: 41
Loan loss provisioning, bank credit and the real economy 31
The number of observations is relatively small as we use annual time
series. The reason for this is that the provisions variable—our chief vari-
able of interest—is only available at an annual frequency. Figure 2.C.2 in
Appendix 2.C shows results from a panel-VAR with four lags estimated
on quarterly data, where we interpolated the provisions series using the
quadratic match average conversion method and the HP output gap de-
rived from quarterly real GDP data.
The IRFs show that the main results remain intact. However, there are
some differences with respect to the magnitude. First, the output gap in-
creases after a positive credit shock; however, the increase is smaller than
the corresponding increase in Figure 2.3 (panel B3). Second, the decrease in
provisioning after a credit shock is smaller (panel A1). This could be due
to the fact that data interpolation does not sufficiently take into account the
volatility of provisions within the year as it smoothens the series between
two observed annual data points.
2.5 Conclusion
In this chapter we have set up a macroeconomic framework including a
banking sector and credit default risk. The banking sector maximizes profits
from lending activities anticipating that a fraction of credit will default in the
future. The solution to the banks’ optimization problem is embedded in a
macroeconomic framework. We estimated the model using a panel-VAR for
12 OECD countries over the last two or three decades to assess the import-
ance of credit default risk. Thereby we used aggregate loan loss provisioning
as a proxy for credit default risk in the banking sector.
Overall, the empirical results are in line with the predictions of the the-
oretical model. First, the results suggest that credit risk (as measured by
loan loss provisioning by banks) is one of the main drivers of business cycle
fluctuations. Specifically, an increase in provisioning decreases bank lending
and economic activity. Second, it appears that banks decrease provisioning
as a percentage of total bank assets as bank lending increases and vice versa.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 42PDF page: 42PDF page: 42PDF page: 42
32 Chapter 2
Hence, during upswings banks take on more risk by building up relatively
low provisions while in a downswing banks build up more loan loss provi-
sions. Third, output is an important determinant of bank lending, more so
than other factors such as interest rates. The sensitivity analysis shows that
our main results remain intact although the magnitude is sometimes weaker
when using a different definition for the output gap or when interpolating
annual provisions data into quarterly data for quarterly estimation.
Our macroeconomic model predictions and empirical findings confirm
the evidence found in the empirical finance literature that loan loss provi-
sions are mostly pro-cyclical and backward-looking. Whereas the literature
found that backward-looking provisioning amplifies credit fluctuations, our
macroeconomic modeling approach links provisioning behavior to the busi-
ness cycle. Our finding that loan loss provisioning has a negative impact
on bank lending and amplifies business cycle volatility, is consistent with
the findings of existing micro-oriented empirical literature, such as Bikker
and Metzemakers (2005) and Laeven and Majnoni (2003). Specifically, the in-
curred loss model, as implemented under International Accounting Stand-
ards (IAS) 39, has been viewed as recognizing impairment losses “too little
and too late” and promoting cyclicality.
One of the policy implications of our findings is that a forward-looking
loan loss provisioning practice rather than a backward-looking one is called
for to avoid pro-cyclicality. Indeed, after the global financial crisis, and sug-
gested by the Financial Stability Board, the G-20 and the Basel Committee
on Banking Supervision initiated a project to replace the incurred loss model
with the expected loss model. This has resulted into the change-over from
the incurred loss model under IAS 39 toward the expected loss model under
International Financial Reporting Standards (IFRS) 9, scheduled to become
effective in 2018 (e.g., Gaston and Song 2014). Under IFRS 9, banks will have
to provision not only for credit losses that have already occurred but also for
losses that are expected in the future. Our findings suggest that under this
new regime, the degree of pro-cyclicality induced by loan loss provisioning
will be considerably mitigated or may even disappear.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 43PDF page: 43PDF page: 43PDF page: 43
Loan loss provisioning, bank credit and the real economy 33
2.A Model solution
The inverse demand function is represented by the following relationship:
ilt = f
(cn
t−k, yt−k, pt−k)
. (2.A.1)
The profit function of banks is denoted as follows:
jt =(
iltϑt − is
t
)cn
t . (2.A.2)
Substituting (2.A.1) into (2.A.2) gives:
jt =[
f(cn
t−k, yt−k, pt−k)
ϑt − ist]
cnt . (2.A.3)
We assume that banks maximize their profits, (2.A.3), with respect to new
credit ∂jt∂cn
t= 0:
f(cn
t−k, yt−k, pt−k)
ϑt − ist + f
′ (cn
t−k, yt−k, pt−k)
cnt = 0,
f(cn
t−k, yt−k, pt−k)=
1ϑt
(ist − f
′ (cn
t−k, yt−k, pt−k)
cnt
). (2.A.4)
Note, (2.A.4) is (2.3) in the main text. The credit payback probability is modeled
according the following equation:
ϑt =ct − Etdt+1
ct=
ct − dt
ct, (2.A.5)
Substituting (2.A.5) into (2.A.4) gives (2.5) in the main text:
f(cn
t−k, yt−k, pt−k)=
ct
ct − dt
(ist − f
′cn(cn
t−k, yt−k, pt−k)
cnt
). (2.A.6)
We can solve (2.5) for new credit in period t:
cnt = −
(ct−dt
ct
)f(cn
t−k, yt−k, pt−k)− is
t
f ′cn
(cn
t−k, yt−k, pt−k) + εc
t, (2.A.7)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 44PDF page: 44PDF page: 44PDF page: 44
34 Chapter 2
where εct denotes a credit shock. Notice, however, that cn
t−k also contains the
term cnt for k = 0. Since we do implicitly assume a functional form for the
loan demand function, we cannot solve for cnt explicitly. Nevertheless, we
can postulate that, disregarding the credit shock term for the moment:
cnt = x1
(dt−k, yt−k, pt−k, is
t−k, ct−k, cnt−k−1
), (2.A.8)
where x1(·) is a function operator. We also know that:
cnt−k−1 = x2
(dt−k−1, yt−k−1, pt−k−1, is
t−k−1, ct−k−1, cnt−k−2
), (2.A.9)
where x2(·) is a function operator. Hence, we can substitute out all cnt−k and
denote cnt as a function x3(·) of the following variables:
cnt = x3
(dt−k, yt−k, pt−k, is
t−k, ct−k)
. (2.A.10)
Substituting (2.A.10) into (2.A.1) gives:
ilt = g
(dt−k, yt−k, pt−k, is
t−k, ct−k)
. (2.A.11)
For convenience we reproduce (2.7) and (2.8) from the main text, respect-
ively:
ct =λct−1 − dt + cnt , 0 < λ < 1, (2.A.12)
dt =(1− ρ)δct−1 + ρdt−1 + εpt , 0 < ρ < 1. (2.A.13)
Substituting (2.A.10) and (2.A.13) in (2.A.12) gives:
ct =λct−1 − (1− ρ)δct−1 + ρdt−1 + εpt +
x3(dt−k, yt−k, pt−k, is
t−k, ct−k)+ εc
t, (2.A.14)
We rewrite (2.A.15) as an implicit function of credit:
ct =h(dt−k, yt−k, pt−k, is
t−k, ct−k)+ εc
t, (2.A.15)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 45PDF page: 45PDF page: 45PDF page: 45
Loan loss provisioning, bank credit and the real economy 35
where h(·) is a function operator. We assume that the credit equation is ad-
ditively separable. Using all the contemporaneous restrictions imposed, the
model can be represented by (2.12).
2.B Variable names and definitions
See Table 2.B.1.
Table 2.B.1. Variable names and definitions.
Variable Notation Source Definition
Inflation πt OECD National Accounts ∆ log of price deflator of privateconsumption (1990 = 100) multipliedby 100%
Short-term ist IMF International Financial Three-month money market interest
interest rate Statistics (IFS) rate (%)Credit ct IMF International Financial ∆ log of bank credit to the private
Statistics (IFS) sector, deflated by the price deflatorof private consumption multipliedby 100%
Provisions dt OECD Bank Profitability First difference of net provisions, i.e.,expense set aside as allowance forbad loans, minus releases, Percentagesof total bank assets
OECD output gap yt OECD Economic Outlook Deviation of actual real GDP fromStatistics, discontinued since potential real GDP as a percentage2009 of potential real GDP
HP output gap yt Own calculations Deviation of actual real GDP frompotential real GDP as a percentageof potential real GDP. Potentialoutput calculated using the Hodrick-Prescott filter on actual output
2.C Robustness checks
See Figs. 2.C.1 and 2.C.2.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 46PDF page: 46PDF page: 46PDF page: 46
36 Chapter 2
Figure 2.C.1. Impulse response functions for provisioning, Aggregate De-mand (AD) and credit shocks
-0.1
0.0
0.1
0.2
0.3
0 3 6
(A1) Provisioning to provisioning
shock
-2.0
-1.0
0.0
1.0
0 3 6
(B1) Output gap to provisioning shock
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0 3 6
(C1) Credit to provisioning shock
-0.5
0.0
0.5
1.0
1.5
0 3 6
(B2) Output gap to AD shock
-0.01
0.00
0.01
0 3 6
(A2) Provisioning to AD shock
-0.1
0.0
0.1
0.2
0 3 6
(C2) Credit to AD shock
-0.12
-0.08
-0.04
0.00
0 3 6
(A3) Provisioning to credit shock
-1.0
-0.5
0.0
0.5
1.0
1.5
0 3 6
(B3) Output gap to credit shock
0.0
2.0
4.0
6.0
0 3 6
(C3) Credit to credit shock
Note: Annual data. HP output gap instead of OECD output gap. Note: 90% confidence in-tervals generated by 1000 Monte-Carlo iterations; periods in years on the horizontal axis.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 47PDF page: 47PDF page: 47PDF page: 47
Loan loss provisioning, bank credit and the real economy 37
Figure 2.C.2. Impulse response functions for Cost Push (CP) and MonetaryPolicy (MP) shocks, quarterly data
-0.02
-0.01
0.00
0.01
0.02
0 3 6 9 12
(A2) Provisioning to AD shock
-0.10
-0.05
0.00
0.05
0.10
0 3 6 9 12
(A3) Provisioning to credit shock
-0.2
-0.1
0.0
0.1
0.2
0.3
0 3 6 9 12
(A1) Provisioning to provisioning
shock
-0.5
0.0
0.5
1.0
1.5
0 3 6 9 12
(B2) Output gap to AD shock
-0.4
-0.2
0.0
0.2
0.4
0 3 6 9 12
(C2) Credit to AD shock
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
0 3 6 9 12
(B3) Output gap to credit shock
0.0
2.0
4.0
6.0
0 3 6 9 12
(C3) Credit to credit shock
-0.6
-0.4
-0.2
0.0
0.2
0.4
0 3 6 9 12
(B1) Output gap to provisioning shock
-0.6
-0.4
-0.2
0.0
0.2
0 3 6 9 12
(C1) Credit to provisioning shock
Note: 90% confidence intervals generated by 1000 Monte-Carlo iterations; periods in quar-ters on the horizontal axis.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 48PDF page: 48PDF page: 48PDF page: 48
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 49PDF page: 49PDF page: 49PDF page: 49
Chapter 3
Credit defaults and bank
capital∗
3.1 Introduction
The global financial crisis demonstrated the importance of banks in determ-
ining macroeconomic fluctuations. In particular, banks started a delever-
aging process and ceased credit supply in response to the materialization of
unexpected credit defaults losses. The decline in credit supply put upward
pressure on lending rates and thereby on firms’ cost of funding. Credit de-
fault losses, however, are sunk costs and should therefore not impact the
banks’ ability to issue new debt and bank equity to finance new credit.
Even though almost a decade has passed since the start of the global finan-
cial crisis, non-performing loans are still a major problem in the euro area,
where firms are largely bank financed. The precise channels and conditions
through which credit default losses affect economic activity for a prolonged
period are, however, largely unknown.
A vast literature has studied the interactions between the financial sector,
the central bank and the real economy. Presumably the most salient line of
literature consists of, among others, Bernanke and Gertler (1989), Kiyotaki
and Moore (1997), Bernanke et al. (1999), Curdia and Woodford (2009, 2010),
∗This chapter is based upon Pool (2016).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 50PDF page: 50PDF page: 50PDF page: 50
40 Chapter 3
Del Negro et al. (2010) and Christiano et al. (2014). These models show the
importance of financial frictions caused by asymmetric information because
they increase the persistence of an adverse shock and amplify the impact
of the shock on the real economy. However, in these papers the financial
sector is often a veil because firms essentially borrow directly from house-
holds. Goodfriend and McCallum (2007), Gerali et al. (2010), Gertler and
Kiyotaki (2010) and Gertler and Karadi (2011) model the impact of bank bal-
ance sheets and leverage constraints on the real economy more explicitly. In
this chapter, we combine the literature on financial frictions and bank lever-
age constraints to account explicitly for the impact of credit default losses
on the bank balance sheet.
This chapter presents a model with an explicit role for credit default
losses on bank balance sheets to explain why credit default losses impact
economic activity. Credit default losses are introduced by augmenting a
standard Smets and Wouters (2003) model with a banking sector as in Gerali
et al. (2010) and default risk as in Bernanke and Gertler (1989) and Bernanke
et al. (1999). In addition, we allow for credit default shocks: a realization
of credit default losses different from expected credit default losses. When
credit default losses are higher than anticipated ex-ante, bank capital deteri-
orates. Banks maximize profits, but are restricted by a leverage (assets-to-
capital ratio) constraint imposed by the regulator.1 We fit the model to euro
area data. Apart from standard macroeconomic data series, we also make
use of financial data on lending rates, deposit rates, NFC-loans, household
deposits and credit spreads (Gilchrist and Mojon, 2017) to identify the credit
default shock.
The estimation results show that an unexpected increase in credit de-
fault losses raises bank leverage. This adversely impacts economic activity
because banks raise lending rates and reduce credit supply. These results
suggest that banks do not or cannot issue new bank equity and, as a con-
sequence, credit supply is constrained by their leverage ratio. The central
1 Shin (2010) shows that profit maximizing banks target leverage ratios because more riskcorresponds to potentially higher returns with limited liability for the bank equity holders.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 51PDF page: 51PDF page: 51PDF page: 51
Credit defaults and bank capital 41
bank decreases the policy rate but, as banks are constrained by their lever-
age ratio, monetary transmission is impeded and lending rates remain high.
In contrast, the deposit rate falls which reduces savings and accordingly the
economic recovery is supported by an increase in household consumption
while investment remains subdued. Inflation falls only marginally because
firms set prices as a mark-up over their marginal costs. While wages de-
cline, funding costs increase which attenuates the decline in inflation. Con-
sistent with this interpretation, the historical shock decomposition shows
that credit default shocks are a major driver of historical fluctuations in out-
put via investment.
The results in this chapter suggest that it is important for the central
bank to identify whether credit demand or credit supply is the culprit of
impeded monetary transmission. If banks are undercapitalized and are re-
luctant to issue new bank equity, lowering the policy rate resembles pushing
a string as the banks’ cost of funding is not the binding constraint and relax-
ing it has little direct impact on credit supply. While decreasing the policy
rate is advantageous as it allows banks to lower the deposit rate and sup-
ports banks to rebuild their capital, accumulating bank capital from retained
earnings is a slow process. The historical decompositions show that credit
default shocks have been an important driver of output and investment fluc-
tuations during the recent global financial crisis. Accordingly, these results
provide a structural explanation for the slow economic recovery following
the global financial crisis despite unprecedented monetary expansion.
As conventional monetary transmission is impeded, we introduce two
alternative (counterfactual) instruments that relax the binding bank capital
constraint directly: a countercyclical capital buffer and a recapitalization.
Benes and Kumhof (2015) show that countercylical capital buffers can com-
plement conventional monetary policy and lead to significant welfare in-
creases. Paries et al. (2011) also show that countercyclical capital buffers can
support monetary policy, but emphasize the importance of operating with
a lengthy implementation schedule to smooth out the transition costs for a
capital constrained banking sector. We find that countercyclical capital buf-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 52PDF page: 52PDF page: 52PDF page: 52
42 Chapter 3
fers can be effective in mitigating the effects of a credit default shock on out-
put and inflation. However, there is a trade-off because banks rebuild their
capital faster without the countercyclical capital buffer. Moreover, counter-
cyclical capital buffers only work when regulatory constraints bind while
market constraints do not. At the height of the global financial crisis, it is
doubtful whether this was the case.
In contrast, in the model a forced recapitalization overcomes this trade-
off problem as rebuilding bank capital is no longer the bank’s choice and it
also mitigates market constraints. A recapitalization can therefore be an ef-
fective instrument to attenuate output and inflation fluctuations when banks
are undercapitalized. Although the degree of effectiveness depends on how
the recapitalization is financed, the results are in general straightforward. As
the household in this model is both the taxpayer, the bank equity holder and
the depositor, converting deposits into bank equity or a government facilit-
ated bank loan has only little negative impact. The claim of households on
the cash flows generated by the bank is after all unaffected. Meanwhile, the
bank leverage constraint relaxes and therefore the recapitalization restores
monetary transmission and enables a faster economic recovery.
The discussion in this chapter focuses on the potential costs of a bank
recapitalization in the short-run. As bank failure is ruled out a priori and
issues like moral hazard that rise in accordance with a recapitalization are
ignored, the long-run consequences of a bank recapitalization are not clear-
cut. Nevertheless, our results do suggest that measures that relax the bank
capital constraint when banks are undercapitalized support the economic
recovery. These results support empirical and theoretical evidence presen-
ted by Berger and Bouwman (2013) and Clerc et al. (2015), respectively, who
show that an increase in bank capital enhances bank performance during a
banking crisis.
The remainder of this chapter is structured as follows. Section 3.2 de-
scribes the model. Section 3.3 discusses the data, the calibration and the
Bayesian methodology. Section 3.4 presents the estimation results. Section
3.5 concludes.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 53PDF page: 53PDF page: 53PDF page: 53
Credit defaults and bank capital 43
3.2 The model
The model contains two types of agents, households denoted by the super-
script H and entrepreneurs denoted by E who operate the firms. The real
side of the model comes from Smets and Wouters (2003). Households and
entrepreneurs interact via a banking sector which competes in a monopol-
istic competitive environment because they offer heterogeneous financial
products, see Gerali et al. (2010). Entrepreneurs borrow to invest in cap-
ital and operate the intermediate firms to produce an intermediate product.
Intermediate firms might default on their loan. Default risk is introduced
by augmenting the model with a Bernanke et al. (1999) financial accelerator
mechanism.
3.2.1 The real side: households, entrepreneurs and firms
Households
The representative household maximizes its expected utility by choosing
consumption, leisure and deposits.2 The inter-temporal utility function is
separable in consumption and leisure:
maxCH
t (i),1−LHt (i),Dt(i)
Et
∞
∑t=0
(βH)tU(CHt (i), LH
t (i)), (3.1)
where
U(CHt (i), LH
t (i)) ≡ ηct
([CH
t (i)− hCHt−1(i)
]1−σc
1− σc−
ηlt[LH
t (i)]1+σh
1 + σh
),
(3.2)
where CHt (i) denotes consumption of agent i at time t, LH
t (i) denotes hours
worked, Dt(i) denotes deposits, βH is the household discount factor, Et is
an expectation operator, σc is the coefficient of relative risk aversion (inverse
2 Here we consider a cashless limit economy, similar to e.g. Smets and Wouters (2007) andGerali et al. (2010). Consequently, the role of liquidity cannot be analyzed as in e.g. Christianoet al. (2005), Goodfriend and McCallum (2007) and Mierau and Mink (2016).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 54PDF page: 54PDF page: 54PDF page: 54
44 Chapter 3
of the inter-temporal elasticity of substitution), σh represents the inverse of
the elasticity of work effort with respect to the real wage, h denotes the
habit parameter, ηct and ηl
t represent a preference shock and a labor sup-
ply shock, respectively. Both shocks follow a stochastic process of the form
ηct = ρcηc
t−1 + εct and ηl
t = ρlηlt−1 + εl
t, 0 < ρb, ρl < 1, where εct and εl
t are
i.i.d. error terms ∼ (µc, σc) and i.i.d.∼ (µl , σl).3 The representative house-
hold maximizes expected utility subject to a series of budget constraints:
CHt (i) + Dt(i) = wtLH
t (i) +1 + rd
t−1
πtDt−1(i) + Πt(i) + (1−ωb)Jt(i)− Tt,
(3.3)
where wt denotes the real household wage Wt/Pt, where Pt denotes the
price level, πt ≡ Pt/Pt−1 is the inflation rate, rdt is the deposit rate such that
(1 + rdt−1)Dt−1(i)/πt denotes real interest income on last period’s deposits,
Πt(i) and (1− ωb)Jt(i) denote real profits from the intermediate firms and
real profits from the banking sector that households’ receive in a lump-sum
fashion (so households are the true owners of the firms and the banks) and
Tt is a lump-sum tax levied by the central bank for the recapitalization of the
banking sector.4
Households are wage setters in the labor market. We adopt a commonly
used wage-adjustment formulation which is a variant of Calvo (1983) pri-
cing. Following Smets and Wouters (2003), we assume that wages can only
change after receiving a random wage signal. The probability of receiving
the signal is equal to 1− ξw. If a household receives a wage signal, it sets a
new wage denoted by Wt. In addition, if no wage signal is received, wages
are indexed partially. If households cannot re-optimize, the wage rate is in-
3 All stochastic shocks are specified as a linear process as the model is eventually linearizedaround its steady state. The non-linear process could be represented by ηt = ρηt−1 + (1−ρ) + εt such that in steady state η∗ = 1.
4 We follow the mainstream approach and assume that actuarially fair priced state-contingent securities exist that insure each household against idiosyncratic variations inlabor and dividend income. Consequently, as in the Arrow-Debreu model individual house-hold income will correspond to aggregate household income.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 55PDF page: 55PDF page: 55PDF page: 55
Credit defaults and bank capital 45
dexed according the following indexation equation:
Wt(i) =(
Pt−1
Pt−2
)γw
Wt−1(i), (3.4)
where γw is the degree of wage indexation. Households maximize utility
(3.1) subject to the budget constraint (3.3) and demand for labor which is
assumed to be represented by:
LHt (i) =
(Wt(i)
Wt
)−(1+λwt )/(λ
wt )
LHt , (3.5)
where λwt = λw + ηw
t determines the wage mark-up, ηwt is a wage cost-push
shock following a stochastic progress: ηwt = ρwηw
t−1 + εwt , 0 < ρw < 1, where
εwt is an i.i.d. error term ∼ (µw, σw). The maximization problems results in
a wage mark-up equation which determines with Equation (3.4) the law of
motion for the wage process, i.e., the so-called New Keynesian wage curve,
see Appendix 3.A.
Entrepreneurs
Entrepreneurs operate the intermediate firms. They invest in real physical
capital Kt at time t which has a nominal price qt and use capital together with
labor to produce. The capital accumulation identity is denoted as follows:
Kt(j) ≡ (1− δk)Kt−1(j) +[
1− ψ
(ηi
t It(j)It−1(j)
)]It(j), (3.6)
where It(j) denotes investment by entrepreneur j, and ψ(·) captures capital
adjustment costs, where ψ(·)′ > 0, ψ(·)′′ < 0. Following Christiano et al.
(2005), we assume that ψ(·) = 0 and ψ′(·) = 0 in steady state, so the ad-
justment costs will only depend on the second-order derivative ψ′′(·). Here
ηit is an investment shock which follows a stochastic process of the form
ηit = ρiηi
t−1 + εit where εi
t is an i.i.d. error term ∼ (µi, σi).
Return to capital is subject to idiosyncratic risk. Ex-post gross return to
capital is given by ωt(j)rkt where ωt(j) is an idiosyncratic disturbance term
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 56PDF page: 56PDF page: 56PDF page: 56
46 Chapter 3
realized by entrepreneur j and rkt is the aggregate gross return on capital av-
eraged over all firms. Similar to Bernanke et al. (1999), we assume that ωt(j)
is independently and identically distributed across entrepreneurs and time
and follows a log-normal distribution with density and cumulative distri-
butions functions denoted by f (ωt(j)) and F(ωt(j)), respectively.
The entrepreneur finances the acquisition of capital with its income in
the previous period and borrows the remaining funds from the banking
sector pledging expected return to capital (Etωt+1(j) rkt+1qtKt(i)) as col-
lateral. Households will not directly finance entrepreneurs because, by as-
sumption, they do not have the means or skills to eliminate all idiosyn-
cratic risk via diversification. If the realization of ωt(j) is below a certain
threshold ωt, the firm defaults because realized return to capital is insuf-
ficient to repay the amount borrowed. The entrepreneur chooses physical
capital, the amount of borrowing and a default threshold by maximizing
the firm’s expected return to capital given a gross non-default lending rate
rbt . The threshold below which the entrepreneur defaults is set according:
rbt Bt(j) = Etωt+1rk
t+1qtKt(i), (3.7)
where Bt(j) denotes the amount borrowed by the entrepreneur. Hence, the
expected default threshold, Etωt+1 = (rbt Bt(j))/(Etrk
t+1qtKt(j)) is de-
termined by the relation between the gross non-default lending rate and the
expected aggregate gross return to capital averaged over all firms rbt /rk
t+1
and a measure related to entrepreneurial leverage, Bt(j)/(qtKt(j)).
Figure 3.1 graphically describes the payoff structure of the loan contract.
If realized return ωt(j) > ωt, the entrepreneur repays the debt rbt Bt(j) and
is entitled to any remaining profits (the area below to dotted 45 degree line
and to the right of ωtrkt qt−1Kt−1 on the x-axis minus the debt repayment
rbt Bt(j)). If ωt(j) < ωt, realized return is insufficient to repay the loan and
the entrepreneur defaults without realizing any profits. The entrepreneur’s
stake in the project is therefore similar to common bank equity. If the entre-
preneur defaults the banks can only recover a fraction (1− µ) of the gross
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 57PDF page: 57PDF page: 57PDF page: 57
Credit defaults and bank capital 47
return, where 0 < µ < 1 denotes the cost of default. In case of default the
banks claim the project’s total return after paying default costs and receive
(1− µ)ωt(j)rkt qt−1Kt−1(j) and lose the remaining part, see Figure 3.1. Thus,
the entrepreneur uses the project’s expected return to capital as collateral.
Figure 3.1. Payoff structure of the loan contract.
Bank return
Total return45
µωtrkt qt−1Kt−1
rbt Bt
ωtrkt qt−1Kt−1
All entrepreneurs who realize an idiosyncratic disturbance term ωt(i) <
ωt default. Using the cumulative distribution function and the law of large
numbers, the fraction of loans that default at period t can be expressed as
F(ωt) =∫ ωt
0 f (ωt)dωt. Using the same notation as in Bernanke et al. (1999)
Γ(ωt) is the share of gross return to capital that goes to the bank:
Γ(ωt) ≡∫ ωt
0ωt f (ωt)dωt + ωt
∫ ∞
ωt
f (ωt)dωt, (3.8)
and µG(ωt) is the share of gross returns that is lost in the bankruptcy pro-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 58PDF page: 58PDF page: 58PDF page: 58
48 Chapter 3
cess:
µG(ωt) ≡ µ∫ ωt
0ωt f (ωt)dωt. (3.9)
Hence, the net share of gross profits that the bank appropriates is equal to
Γ(ωt)− µG(ωt). The financial intermediary has opportunity costs denoted
by the riskless gross rate of return rt, which is the interest rate set by the cent-
ral bank. However, different from the financial intermediaries in Bernanke
et al. (1999), the banking sector might be balance sheet constrained. When
the bank is balance sheet constrained, its appropriate opportunity costs are
no longer the risk-free rate, but a higher lending interest rate, rwbt . The bank
will therefore only agree with the loan contract if it receives on average at
least this lending rate.
Entrepreneurs’ optimization problem
Entrepreneurs only care for consumption. They maximize the discounted
sum of expected future utility, where the utility function takes a logarithmic
form, by choosing consumption, capital, loans, the default threshold value,
the capital utilization rate, and labor input (CEt (j), Kt(j), Bt(j), ωt+1(j), ut(j)
and LHt (j)):
maxCE
t ,Kt,Bt,ωt+1,ut,LHt
Et
∞
∑t=0
(βE)t ln
CEt − hCE
t−1
subject to their budget constraint:
Bt(j) =1 + rb
t−1 [1− Γ(ωt(j))]πt
Bt−1(j)−
Et
[1− Γ(ωt+1(j))]
[rk
t+1ut − ψ(ut)]
Kt(j)+
It(j) + wtLHt (j) + CE
t (j), (3.10)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 59PDF page: 59PDF page: 59PDF page: 59
Credit defaults and bank capital 49
the capital accumulation identity (3.6) and the participation constraint of the
banks:
Et
[Γ(ωt+1(j))− µG(ωt+1(j))
]rk
t+1utqtKt(j)≥ rwb
tEt πt+1
Bt(j),
(3.11)
where ψ(ut(j))Kt−1(j) is the real cost of setting a capital utilization rate
equal to ut and δk is the depreciation rate of physical capital. Inequality
(3.11) states that banks invest only if they expect the project’s return to be
higher than their opportunity costs of supplying credit denoted by rwbt . In
the steady state this opportunity cost is equal to the risk-free interest rate rt.
Banks, however, might be constrained by the amount of capital they have. In
that case, capital requirements limit the bank to supply more credit which
raises the lending rate. As the right hand side of (3.11) increases the left
hand side must increase as well. That is, if credit conditions tighten firms
need more capital, a higher rate of return, higher capital prices or a lower
default probability to obtain a similar amount of funding.
Intermediary firms and retailers
Intermediate firm j produces a unique variety of a wholesale good Yt(j)
according to the following Cobb-Douglas production function:
Yt(j) = ηat AtKt(j)αLH
t (j)1−α, (3.12)
where At is a Hicks-neutral technology parameter and ηat is a technology
shock which follows a stochastic progress of the form ηat = ρaηa
t−1 + εat where
εat is an i.i.d. error term ∼ (µa, σa). The firm produces the wholesale good
using capital and labor hired from the household sector at a real wage rate
wt.
Intermediate firms sell their wholesale products to retailers who trans-
form the intermediate product in a homogeneous product by application of
a CES production function. The introduction of intermediate and retail firms
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 60PDF page: 60PDF page: 60PDF page: 60
50 Chapter 3
is merely a mechanical device to keep the model analytically tractable. The
CES production function is represented by:
Yt =
[∫ 1
0Yt(j)1−1/λ
pt di]1/(1−1/λ
pt )
, (3.13)
where λpt = λp + ηπ
t denotes the price mark-up, ηπt is a cost-push shock and
follows a stochastic progress of the form ηπt = ρπηπ
t−1 + επt where επ
t is an
i.i.d. error term ∼ (µπ, σπ). Retailers minimize costs,∫ ∞
0 Pt(j)Yt(j), subject
to the CES production function, (3.13). The solution to the retailers’ optimiz-
ation problem defines how the price and output of intermediary j, Pt(j) and
Yt(j), respectively, relate to aggregate prices and aggregate output Pt and
Yt, respectively. We assume that retailers compete in a perfectly competitive
market which implies that prices can be rewritten as:
Pt =
[∫ 1
0Pt(j)1−λ
pt di]1/(1−λ
pt )
. (3.14)
We adopt Calvo (1983) pricing as retailers can only change their price after
receiving a random price change signal. The exogenous probability of re-
ceiving the price signal is equal to (1− ξ p). If a retailer receives a price sig-
nal, she sets a new price denoted by Pt. If a retailer does not receive a price
signal, we allow for partial indexation. Partial indexation is done in a way
similar to wage indexation:
Pt(i) =(
Pt−1
Pt−2
)γp
Pt−1(i), (3.15)
where γp is an indexation parameter for non-optimizing firms. As a con-
sequence, prices in the model are sticky.
3.2.2 The financial side: banks
Banks act as intermediaries for all financial transactions between house-
holds and entrepreneurs. Households save deposits to smooth consumption
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 61PDF page: 61PDF page: 61PDF page: 61
Credit defaults and bank capital 51
over time and entrepreneurs borrow to finance production. Following Gerali
et al. (2010) the banking sector is modeled as a collaboration between three
branches, i.e., a bank holding company and two retail branches: a fund-
ing branch and a lending branch. This approach ensures tractability of the
model. The two retail branches are responsible for the collection of deposits
and the allocation of loans and set interest rates in a monopolistic competit-
ive fashion. The bank holding company manages the capital position of the
banking entity.
Loan and deposit demand
Banks and borrowers often engage in long-term relationships which are vul-
nerable to asymmetric information problems. Due to this market character-
ization switching banks is considered costly, because the lender has to alloc-
ate costs to screen potential new borrowers and the borrower has to signal
creditworthiness to the new lender. The presence of switching costs due to
asymmetric information is often mentioned as a reason for market power in
the banking sector, see Diamond and Dybvig (1983), Greenbaum et al. (1989)
and Sharpe (1990).5
Market power in the banking sector is modeled by application of a Dixit-
Stiglitz framework. Each bank produces a unique variety of loans Bt(`) and
deposits Dt(`). Accordingly, loan demand by entrepreneurs and deposit de-
mand by households at bank ` are given by, see Appendix 3.A for the deriv-
ations:
Bt(`) =
(rb
t (`)
rbt
)µbt
Bt, (3.16)
5 The market structure within the banking sector is also an often cited source of marketpower. Berger et al. (2004) link market concentration to market power and the interest ratesetting behavior of banks. They find evidence that high market concentration in the bank-ing sector increases market power of banks. Other studies report limited contestability andregulatory restrictions as a source of market power, e.g. Demirguc-Kunt et al. (2004). Severalempirical papers confirm the presence of market power in the banking sector, see Bergeret al. (2004) and Degryse and Ongena (2008) for a discussion.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 62PDF page: 62PDF page: 62PDF page: 62
52 Chapter 3
Dt(`) =
(rd
t (`)
rdt
)µdt
Dt, (3.17)
where µbt = µb + ηb
t and µdt = µd + ηd
t denote the elasticity of substitution
for loans and deposits, respectively and ηbt and ηd
t are a loan demand and
deposit demand shock following a stochastic process equal to ηbt = ρbηb
t−1 +
εbt and ηd
t = ρdηdt−1 + εd
t where εbt and εd
t are i.i.d. error terms ∼ (µb, σb) and
i.i.d.∼ (µd, σd).
Bank holding company
The bank holding company operates under perfect competition and com-
bines bank capital Kbt (`) and deposits Dt(`) on the liability side and sup-
plies loans Bt(`) on the asset side to maximize its profits. Moreover, Vrt (`)
denotes the realized (superscript r) level of losses from credit defaults.
The bank holding company expects each period that a number of en-
trepreneurs default on their loan repayment because they realize an idio-
syncratic return that is too low to repay the loan. If an entrepreneur de-
faults, the bank only receives the residual claim net of monitoring costs,
(1− µ)ωt(j)rkt qt−1Kt−1(j), which happens if ωt(j) < ωt. Aggregating over
all entrepreneurs that default gives the aggregate return on loans that de-
fault: (1− µ)∫ ωt
0 ωt f (ωt)dωtrkt qt−1Kt−1. The default probability is equal to
F(ωt) and accordingly the realized amount of bank losses in period t is de-
termined by:
Vrt (`) =
(F(ωt)rb
t (`)Bt−1(`)− (1− µ)∫ ωt
0ωt f (ωt)dωtrk
t qt−1Kt−1
)ηv
t ,
(3.18)
where ηvt denotes a credit default shock. The credit default shock follows a
stochastic process ηvt = ρvηv
t−1 + εvt where εv
t is an i.i.d. error term∼ (µv, σv).
Intuitively ηvt describes the deviation of realized from anticipated losses.
Bank holding companies predict future losses based on historical informa-
tion assuming a log-normal distribution for ωt. However, during a finan-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 63PDF page: 63PDF page: 63PDF page: 63
Credit defaults and bank capital 53
cial crisis the log-normal distribution or the historical default losses may
not be representative for the actual default process. Consequently, actual
losses might differ from expected losses. A higher level of expected losses
increases provisioning by banks for future credit default losses which lowers
the amount of funds available for loans and raises lending rates today. A
realization of credit default losses higher than anticipated ex-ante deterior-
ates bank capital. Whereas both channels are present in the model, the credit
default shock only focuses on the latter.6
Bank holding companies determine expected default losses in the next
period, EtVrt+1(`), and reserve the equivalent today denoted by Ve
t (`) (su-
perscript e). Hence, Vet (`) = EtVr
t+1(`) and the amount of funds reserved
for future losses is based on today’s information set. Each bank holding
company has a balance sheet constraint which is given by:
Bt(`) = Dt(`) + Kbt (`). (3.19)
Bank capital, or bank equity, represents a residual claim on the bank’s profits
after having payed the debtors, operational costs and dividends:
Kbt (`) = (1− δb)Bt−1(`)− Dt−1(`) + ωb Jb
t (`) + Tt, (3.20)
where Jbt denotes overall bank profits of the retail banks and bank holding
company, (1−ωb) denotes the dividend payout ratio of the bank, δb denotes
resources used to manage the assets and Tt captures the effect of a bank
recapitalization. The precise bank recapitalization rule is specified below.
All retained earnings accumulate to bank capital and are reinvested the next
period. Using (3.19) and (3.20) it is possible to write bank capital as:
Kbt (`) = Kb
t−1(`)− δbBt−1(`) + ωb Jbt (`) + Tt. (3.21)
6 This modeling assumption is in line with existing evidence in the empirical literature onloan loss provisioning, see e.g., Laeven and Majnoni (2003) and Bikker and Metzemakers(2005) who show that credit default losses impact bank capital directly because banks do notprovision extra in good times to build a buffer for bad times.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 64PDF page: 64PDF page: 64PDF page: 64
54 Chapter 3
Moreover, the bank holding company faces a cost, denoted by the parameter
κw, whenever the value of the capital-to-assets ratio Kbt /Bt (the inverse of
the bank’s leverage ratio) deviates from the optimal capital-to-asset ratio ν.
If the capital-to-asset ratio is below the target ratio, banks are leveraged too
high and might incur insolvency costs. Insolvency costs are not explicitly
modeled because banks cannot default in our model. For this reason, we
capture the insolvency costs via the parameter κw. Without this constraint
banks would have an incentive to increase leverage indefinitely.
If the capital-to-asset ratio is above the target ratio, banks are not max-
imizing their profits as a lower ratio (higher bank leverage) would result
in higher returns. In this case, the bank’s portfolio is not on the Markowitz
frontier because banks could increase the expected return without increas-
ing volatility. In a Modigliani-Miller world a lower leverage ratio corres-
ponds to a lower risk level such that investors in the bank would require
a lower return. In this model banks cannot default. Consequently, a lower
leverage ratio does not correspond to a lower risk level which, on itself, does
not violate Modigliani-Miller. Nonetheless, banks have an incentive to max-
imize their leverage ratio and hold zero bank equity when the costs of bank
equity are higher than the cost of debt and therefore gravitate towards the
highest leverage ratio possible.
Shin (2010) shows that even if banks face bankruptcy risk, leverage tar-
geting is optimal as long as the maximum leverage ratio required by the
market absent regulation is higher than the maximum leverage ratio re-
quired by the central bank. In the real world banks have implicit govern-
ment guarantees if they are perceived “too big to fail” and depositors are
protected via deposit insurance systems. As the market is aware of these im-
plicit guarantees, leverage ratios required by the market often do not bind
because of regulatory constraints and banks have an incentive to raise their
leverage ratio.
The model does not allow banks to sell loans to other financial market
participants thereby lowering leverage contemporaneously. This assump-
tion reflects the situation in the wake of the financial crisis during which all
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 65PDF page: 65PDF page: 65PDF page: 65
Credit defaults and bank capital 55
banks tried to sell their assets simultaneously. The interbank markets dried
up completely and prevented banks to sell their loan portfolio, usually in
the form of asset backed securities, to other market participants.
The model also does not allow banks to issue new shares. Issuing bank
equity is in general considered to be too expensive by the current share-
holders and only done by a few banks. We do not model the bank decision
to issue new equity explicitly, i.e., the model is agnostic about why the
Modigliani-Miller irrelevance proposition is violated. However, with impli-
cit government guarantees, banks can continue to operate even though the
market value of their debt is larger than the market value of their assets,
i.e., banks are effectively bankrupt. The government guarantee protects the
risk-free claim of the depositor and, as a consequence, the market value of
bank equity can still be positive. In this case, issuing new equity is indeed
expensive as it lowers the value of the implicit government guarantee and
thereby violates the Modigliani-Miller irrelevance proposition.
The bank holding company maximizes the discounted sum of the expec-
ted future cash flows by choosing the appropriate loan and deposit levels
subject to the balance sheet constraint:
maxBt(`),Dt(`)
Et
∞
∑t=0
Λpt
[(1 + rwb
t (`))Bt(`)− Bt+1(`) + Dt+1(`)−
(1 + rwdt (`))Dt(`) + ∆Kb
t+1(`)−
κw
2
(Kb
t (`)
Bt(`)− ν
)2
Kbt (`)
],
subject to Bt(`) = Dt(`) + Kbt (`), (3.22)
where Λpt is the stochastic discount rate of the bank holding company, rwd
t
and rwbt denote the deposit rate and the loan rate charged by the bank hold-
ing company to the corresponding retail branches, respectively, and ∆Kbt+1
denotes the change in bank capital. As the model is estimated in linear form,
the leverage adjustment costs are completely specified by the parameter κw.
The quadratic term is postulated for mathematical convenience. Using the
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 66PDF page: 66PDF page: 66PDF page: 66
56 Chapter 3
balance sheet constraint at time t and t + 1 in the objective function, (3.22),
can be rewritten as:
maxBt(`),Dt(`)
rwb
t (`)Bt(`)− rwdt (`)Dt(`)−
κw
2
(Kb
t (`)
Bt(`)− ν
)2
Kbt (`)
. (3.23)
The FOCs link the bank holding rates on loans and on deposits to the degree
of leverage Bt/Kbt . As banks are always solvent, they are never financing
constrained and can always borrow from the central bank at rate rt. How-
ever, they cannot use these funds to increase their loan portfolio unlimitedly
as they are constrained by their leverage ratio. Arbitrage opportunities en-
sure that rwdt = rt. Using these results the FOCs can be rewritten as:
st(`) ≡ rwbt (`)− rt = −κw
(Kb
t (`)
Bt(`)− ν
)(Kb
t (`)
Bt(`)
)2
ηst , (3.24)
where ηst denotes a spread shock which follows a stochastic process ηs
t =
ρsηst−1 + εs
t where εst is an i.i.d. error term ∼ (µs, σs). Equation (3.24) links
the rate of the bank holding company to the central bank interest rate and
to bank leverage. The difference between the bank holding rate and the
risk-free rate, st, is determined by bank leverage. When expected defaults
increase, expected profits and future bank capital decline. Banks decrease
credit supply to ensure that in expectation leverage remains constant and
the spread is equal to zero. However, when the realization of defaults turns
out to be higher than anticipated, profits decline which increases leverage
contemporaneously. As a consequence, the rate set by the bank holding com-
pany increases and st can be interpreted as a credit spread that increases
(decreases) due to unanticipated firm defaults (survivals), because absent
unexpected defaults it would be constant and equal to zero.
In Bernanke et al. (1999) an increase in ωt is accounted for via an increase
of the credit spread. All losses that materialize precipitate on the real side of
the economy and are accounted for in the goods market equilibrium. While
an amplification of the downturn can be expected, leverage or balance sheet
constraints do not play a significant role. In this chapter, an increase in the
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 67PDF page: 67PDF page: 67PDF page: 67
Credit defaults and bank capital 57
default thresholds not only increases credit spreads, but also actual defaults
which deteriorates bank capital of leveraged banks.
Retail branches
The retail branches are monopolistic competitors on both the loan and de-
posit markets. Both the lending branch as well as the funding branch pro-
duce a differentiated product and have some market power.
The lending branch maximizes its profits by lending to entrepreneurs
while financing these lending activities by borrowing from the bank hold-
ing company at rate rwbt . The lending branch maximizes its profits by choos-
ing the appropriate lending rate rbt facing quadratic interest rate adjustment
costs denoted by the parameter κb:
maxrb
t (`)
∞
∑t=0
Λpt
(rbt (`)− rwb
t (`))Bt(`)−κb
2
(rb
t (`)
rbt−1(`)
− 1
)2
rbt Bt
, (3.25)
subject to the loan demand schedule, Equation (3.16). The interest rate ad-
justment costs are introduced to mimic the empirical evidence of a slug-
gish lending rate rate pass-through (see, for example, Sørensen and Werner
(2006)).
Similarly to the lending branch, the funding branch of bank j collects de-
posits Dt(j) from households and passes these to the bank holding company
which compensates them at rate rt = rwdt (the interest rate set by the cent-
ral bank). In addition, the funding branch faces quadratic interest rate ad-
justment costs which are denoted by the parameter κd. The funding branch
maximization problem becomes:
maxrd
t (`)
∞
∑t=0
Λpt
(rt − rdt (`))Dt(`)−
κd
2
(rd
t (`)
rdt−1(`)
− 1
)2
rdt Dt
, (3.26)
subject to deposit demand, Equation (3.17). Interest rate adjustment costs
are introduced also for the funding branch to mimic the empirical evidence
of a sluggish deposit rate pass-through (Sørensen and Werner, 2006).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 68PDF page: 68PDF page: 68PDF page: 68
58 Chapter 3
Combined profits of the bank holding company, and the lending and
funding branches are equal to:
Jbt = rb
t Bt − rdt Dt −
κw
2
(qtKb
tBt− ν
)2
qtKbt −VE
t+1 + VEt −Vr
t − Cbt ,
(3.27)
where Cbt ≡
κb2
(rb
t (`)
rbt−1(`)
− 1)2
rbt Bt +
κd2
(rd
t (`)
rdt−1(`)
− 1)2
rdt Dt are the adjustment
costs for changing the interest rates at the retail level. Hence, bank profits
are determined by interest income on loans minus interest expenses on de-
posits, deviations from the optimal capital-to-asset ratio, anticipated default
losses, unanticipated default losses, and adjustment costs for changing in-
terest rates.
Aggregation and equilibrium
The goods market is in equilibrium if production equals consumption plus
the resources absorbed in the production of capital:
Yt = Ct + It + ψ(ut)Kt−1, (3.28)
where Ct ≡ CHt + CE
t . The rental market for capital is in equilibrium when
the demand for capital by entrepreneurs equals supply by capital produ-
cers: Kt =∫ 1
0 Kt(i)di. The labor market is in equilibrium when labor de-
mand by entrepreneurs equals labor supply of households and entrepren-
eurs: Lt =∫ 1
0 Lt(i)di. Finally, a conventional Taylor rule is postulated to
close the model:
(1 + rt) =(1 + r)(1−ρr)(1 + rt−1)ρr
(πtEtπt+1
π∗
)δπ(1−δr)
×(Yt
Yt−1
)δy(1−δr)
εmt , (3.29)
where εmt is a monetary policy shock which follows an AR(1) process εm
t =
ρmεmt−1 + ηm
t and ηmt is an i.i.d. error term ∼ (µm, σm). Equation (3.29) as-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 69PDF page: 69PDF page: 69PDF page: 69
Credit defaults and bank capital 59
sumes that the central bank has two objectives, a stable (expected) inflation
rate around the steady state inflation rate π∗ and a stable growth path de-
scribed by the deviation of Yt from Yt−1. Moreover, we assume interest rate
smoothing by the central bank and include the lagged policy rate.
For the empirical analysis in Section 3.3 we linearize the model around
the non-stochastic steady state, see Appendix 3.A for details. Throughout
the remainder of this chapter a lower case variable with a hat denotes a log-
linearized variable. It is convenient to summarize the model using matrix
notation. The reduced form of the model can be represented as:
Γ0EtZt+1 = Γ1Zt + Γ2Zt−1 + Υ0Etηt+1+ Υ1ηt, (3.30)
where Γ0, Γ1 and Γ2 are respectively the coefficient matrices specifying the
response of each observable variable at time t + 1 to all variables at time t +
1, t and t− 1, Zt =[yt, ct, it, wt, πt, bt, dt, rt, rb
t , rdt , st,
]′is a vector of observed
variables, Υ0 and Υ1 are coefficient matrices specifying the response of each
variable in Zt+1 to the vectors Etηt+1 and ηt =[ηa
t , ηvt , ηm
t , ηlt , ηi
t, ηπt , ηd
t , ηbt ,
ηct , ηi
t, ηqt]′, respectively, which contain all shocks.
3.2.3 Bank recapitalizations and countercyclical buffers
Since the global financial crisis, central banks have a wide range of instru-
ments at their disposal to alleviate stress in the banking sector. Besides, at
the height of the financial crisis, governments of advanced economies fre-
quently recapitalized banks as an emergency measure. In the current after-
math of the crisis, policy makers are in search for more structural solutions
that could potentially prevent bank recapitalizations in the future given
the potentially damaging side effects in the form of moral hazard. Among
these instruments is a countercyclical capital buffer which suggests that the
central bank tightens capital constraints during a boom and eases capital
constraints during a downturn. Bank capital constraints might induce pro-
cyclical bank behavior and countercyclical capital buffers are suggested to
alleviate this pro-cyclical nature.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 70PDF page: 70PDF page: 70PDF page: 70
60 Chapter 3
Following Paries et al. (2011) we focus on the joint determination of a
monetary policy rule and a macroprudential policy rule. We abstract from
welfare calculations and specify an ad hoc macroprudential policy rule. In
practice, countercyclical capital buffers and bank recapitalizations have a
binary structure, i.e., either the countercyclical capital buffer is activated or
the bank is recapitalized, or not. Here an endogenous rule is introduced
which is not binary but continuous. First, we introduce an endogenous policy
rule that is contingent on the ratio of credit over output which is the most
commonly used trigger variable that activates, e.g., countercyclical capital
buffers:
Myt =
(Bt
Yt− B∗
Y∗
)$y
, (3.31)
where B∗/Y∗ is the steady-state credit-to-output ratio and $y is a policy
parameter which denotes the degree to which the countercyclical capital
buffer or recapitalization is affected by changes in the credit-to-output ratio.
So, Myt is activated (My
t 6= 0) when credit-to-output differs from steady state
credit-to-output.
The credit-to-ouput ratio has been chosen as a trigger variable to activ-
ate the countercyclical buffer because it signals the economy’s ability to re-
pay its debt. A high credit-to-output ratio is informative about future credit
default losses because it might signal, for example, a bubble as credit has
expanded but output has not. The model present here, however, has no role
for inflationary bubbles. As credit is used to acquire capital for production,
credit and output grow often in accordance. Also in reality it is hard to dis-
tinguish between a bubble and an increase in credit supported by funda-
mentals. It is therefore informative to examine a more direct instrument, a
policy rule that is contingent on bank capital itself:
Mkt =
(Kb
t − Kb∗)$k
, (3.32)
where Kb∗ denotes steady state bank leverage, and $k is a policy parameter
which denotes the degree to which the countercyclical capital buffer or re-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 71PDF page: 71PDF page: 71PDF page: 71
Credit defaults and bank capital 61
capitalization is affected by changes in the leverage ratio. So, Mkt is activated
(Mkt 6= 0) when bank leverage differs from steady state bank leverage.
Finally, we assume that banks do not take into account the incremental
impact they have on the countercyclical capital buffer or the recapitaliza-
tion, i.e., no moral hazard is introduced in the model. The rationale for this
assumption is that each individual bank has only marginal influence on the
aggregate development of the credit-to-output ratio or the aggregate lever-
age ratio whereas the rules are only implemented at the aggregate level.
Countercyclical buffers
In reality, countercyclical capital buffers are either off and no capital sur-
charge (undercharge) is required (allowed), or they are activated and banks
are required (allowed) to hold (release), e.g., an extra percent of capital re-
lative to their assets. Countercyclical capital buffers have only an effect on
bank leverage when required bank leverage after activating the countercyc-
lical capital buffer is below the leverage ratio required by the market. Spe-
cifically, in a downturn the central bank may decide that capital buffers are
allowed to fall below the regulatory capital requirement, i.e., bank leverage
may increase. Yet, if the market requires a lower leverage ratio for solvency
reasons, an increase in the bank’s funding costs will force the bank to de-
crease leverage to a level required by the market, see for example Clerc et al.
(2015). In this case the capital requirement set by the central bank is not
binding and has no effect on credit supply.
For this reason, we might postulate that the central bank introduces a
steady state leverage ratio significantly below the market requirement in a
financial crisis. Accordingly, even after activating the countercyclical capital
buffer, the leverage ratio requirement of the central bank is still below the
market requirement. Accordingly, the leverage constraint set by the central
bank always binds, while the constraint imposed by the market never binds.
Also in practice, leverage requirements must be set at a relatively low level
to ensure that the countercyclical capital buffer binds even if financial con-
ditions deteriorate. If not, the countercyclical capital buffer has no effect as
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 72PDF page: 72PDF page: 72PDF page: 72
62 Chapter 3
it is not binding.
Alternatively, the government could offer an explicit or implicit guar-
antee. This guarantee is effective if it does not jeopardize the solvency of
the government as this would increase lending rates and thereby crowd out
private investment. If the government guarantee is credible, and it does not
affect the solvency of the government, bank leverage could increase dur-
ing a downturn because the amount of bank equity required by the mar-
ket decreases. The model presented here is consistent with both interpret-
ations and simply assumes that the regulatory requirement after activating
the countercyclical capital buffer binds.
Formally, the bank leverage constraint is relaxed in the following way
when the countercyclical capital buffer is activated:
rwbt = rt − κw
(Kb
tBt− (ν + Mt)
)(Kb
tBt
)2
ηst , (3.33)
where Mt can be either Myt or Mk
t depending on which activation rule the
central bank adheres to.
Bank recapitalizations
A bank recapitalization could be designed in countless ways as a priori it
is not clear how the recapitalization should be financed. For example, the
recapitalization could be specified as a mandatory bank equity issuance, a
bail-in or a tax-financed bail-out. If the government is solvent, it could also
issue new debt to finance the bank recapitalization without affecting lend-
ing rates and private investment much. This situation might be particularly
relevant in times of a liquidity trap when the costs of debt-financed fiscal
expansions are suggested to be low (Eggertsson and Krugman, 2012). In the
model, the government could in this case recapitalize the banks and reduce
output and inflation fluctuations to zero without incurring any costs.
To analyze a less trivial trade-off, we focus on a costly recapitalization.
As the patient household in the model is both the tax-payer, bank share-
holder and depositor, any form of recapitalization is payed for by the house-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 73PDF page: 73PDF page: 73PDF page: 73
Credit defaults and bank capital 63
hold. If the banks issues equity or if the government recapitalizes the banks
with a loan, the patient households could loose in terms of current and/or
future consumption. Nevertheless, if Modigliani-Miller holds, the banks’
total financing costs are unaffected by their financial structure. Households
effectively convert deposits into bank equity without changing their expec-
ted claim on the banks’ future cash flow and are therefore in a risk neut-
ral setting unaffected. Households, however, might value bank equity less
than deposits for reasons of liquidity which would generate a violation of
Modigliani-Miller, see e.g. Stein (2012). For simplicity we therefore assume
that the household receives a return of zero on bank equity or a loan provided
by the government. Moreover, we analyze the case where the loan to the
bank is offered as a gift.
If the government decides on a recapitalization rather than activating
the countercyclical capital buffer, the tax Tt is set equal to the endogenous
buffer Myt or Mkb
t . In this case the government directly injects the bank with
additional capital as specified in Equation (3.21). It is therefore important to
distinguish between a countercyclical capital buffer and a recapitalization
as the former does not need funding if credible while the latter must be
financed explicitly. The taxed levied on the households is given by:
Tt = rdt Mt. (3.34)
Hence, households loose the deposit rate they would otherwise receive on
their deposits, but the principle claim on the bank remains the same. As
banks are leveraged and households hold both bank equity and deposits, a
one percent increase in bank capital yields a ν/(1− ν) percent reduction in
household deposits where ν is the leverage ratio.
3.3 Empirical methodology
In this section we specify the parameter values and the estimation strategy.
The set of parameters is divided into two partitions. Partition one contains
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 74PDF page: 74PDF page: 74PDF page: 74
64 Chapter 3
parameters that control the steady state. This set of parameters entails con-
ventional parameters for which the values are taken from the literature. The
other parameter values in this set entail steady state averages which are
calibrated to reproduce steady state averages of the data set. Partition two
contains parameters that are estimated using a Bayesian estimation proced-
ure.
3.3.1 Calibrated parameters
The parameters that are conventional are categorized in partition one and
presented in the first column of Table 3.1. Their values are taken from the
literature. First, βE = 0.975 and βH = 0.994 which are also used by Gerali
et al. (2010). The share of capital in the production function α = 0.3; the de-
preciation rate of physical capital δk = 0.025; the habit parameter h = 0.7;
the coefficient of relative risk aversion σc = 1; the inverse of the elasticity
of work effort σl = 2 (see e.g., Smets and Wouters (2003) and Christiano
et al. (2005)). We set the share of inelastic entrepreneurial labor in produc-
tion Ω = 0.01 which is similar to Bernanke et al. (1999) and ensures that
entrepreneurial income has no effect on the results. The inverse of the elasti-
city of capital utilization Ψ ≡ ψ′/ψ
′′= 0.25.
The probability of default F(ω) = 3%. Bernanke et al. (1999) base this
value loosely on the United States historical average. We assume a similar
value for the euro area but experimented with higher and lower values.
These experiments showed that the model outcome is rather insensitive to
the default threshold value because it is not the steady state value that is
important but changes in the default probability and the mismatch between
what is expected and what is realized. F(ω) = 3% pins down the entrepren-
eurial physical capital-to-loan ratio (K/B) at 0.4, which is close to the his-
torical average capital-to-loan ratio of entrepreneurs.7 In this model there is
7 Assuming a log normal distribution ∼ N (0, 1) implies f (ω) ≈ 0.446 and ω ≈ 0.152.Optimization and linearization of the system results in first and second order derivativesw.r.t. Γ(ω) and µG(ω) where Γ(ωt) = F(ω) 1
2 ω2 + ω[1 − F(ω)], µG(ωt) = µF(ω) 12 ω2,
Γ′(ω) = 1 − F(ω), Γ
′′(ω) = − f (ω), G
′(ω) = µω f (ω) and G
′′(ω) = µ( f (ω) + ω f
′(ω))
which are all functions of the steady state default probability ω.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 75PDF page: 75PDF page: 75PDF page: 75
Credit defaults and bank capital 65
by assumption no dividend payout ωb = 1. Assuming a positive dividend
payout has no qualitative effect on the results.
The parameters representing steady state averages are calibrated based
on their historical averages. The fractions C/Y and I/Y are calibrated from
of the data series and incorporate government consumption and govern-
ment investment; wL/D is the labor income to savings ratio and is set equal
to 0.23, roughly the historical average of the data series. Following Gerali
et al. (2010) the share of entrepreneurs in the economy CE/C = 0.2. The
fractions concerning bank profit, rbB/Jb, rdD/Jb, and V/Jb are derived from
estimated parameters and therefore not calibrated, see Appendix 3.A.
Table 3.1. Calibrated parameters
Parameters Description Value
βE Discount factor entrepreneurs 0.975βH Discount factor households 0.994α Share of capital in the production function 0.300h The household habit parameter 0.700σc Relative risk aversion households 1.000σl Inverse of the elasticity of work effort 2.000δk Depreciation rate of physical capital 0.025Ω Share of households in composite labor factor 0.990Ψ = ψ
′(1)/ψ
′′(1) Capital utilization 0.250
F(ω) Probability of default 0.030C/Y Consumption to output ratio 0.780I/Y Investment to output ratio 0.220wL/D Households savings quote 0.230CE/B Consumption borrowing ratio 0.230K/B Inverse of loan to value ratio 0.400
3.3.2 Data
The model is estimated for the euro area for the period 2000:Q1-2016:Q2. The
dataset contains real economic variables: output, consumption, investment,
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 76PDF page: 76PDF page: 76PDF page: 76
66 Chapter 3
hours of work, wages, outstanding loans to firms and outstanding deposits;
and price and interest variables: inflation, nominal policy rate, nominal in-
terest rate on loans, nominal interest rate on deposits and credit spreads (see
Appendix 3.B). The real economic variables are de-trended and expressed
as log-deviations from their trend. The trend value of the variables is con-
structed using the HP-filter and a smoothing parameter equal to 1600. The
prices and interest rate variables are expressed as absolute deviations from
the sample mean. Figure 3.2 shows the resulting time series.
The difference between the lending rate set by bank holding company
rwt and the policy rate rt is interpreted as a credit spread. We take credit
spread data of European firms constructed by Gilchrist and Mojon (2017)
who use individual firm level securities data to construct security-specific
credit spreads which are aggregated for the euro area to construct aggreg-
ated credit spread indicators. As the aggregate indicator is constructed from
micro level data, the spreads are, according to the authors, more informative
about aggregate credit spreads than aggregate approximations. The credit
spread data is added to be able to identify the default shock.
All real variables presented in Figure 3.2 show more or less the same pat-
tern. Output, consumption, investment, hours and wages peak before the
global financial crisis and start to decline during the global financial crisis.
Although they return swiftly to pre-crisis levels, the resurrection might also
be a characteristic of the HP-filter applied to the data series as the estim-
ated trend is not insensitive to the financial crisis. Loans show also a sharp
decline during the financial crisis and return quickly to pre-crisis levels. De-
posits show the inverse pattern. Before the global financial crisis they are at
their lowest point after which they rise steeply. The policy rate, the loan rate
and to a lesser extent the deposit rate appear to be trending downward over
the sample. Credit spreads, in contrast, appear to be trending upward; they
hit a minimum just before the start of the global financial crisis at about
5 percent points below their mean, while during the crisis they appear to
peak. The inflation rate is relatively stable over time reflecting perhaps ef-
fective monetary policy over the sample period.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 77PDF page: 77PDF page: 77PDF page: 77
Credit defaults and bank capital 67
3.3.3 Estimation
The second partition of parameters, containing the less conventional para-
meters, are estimated using the Bayesian estimation algorithm in Dynare.
The estimated parameters are presented in Table 3.2.
We assume that all the structural parameters follow a gamma distribu-
tion. The standard deviations are set relatively large to give the data the
opportunity to determine the value of the parameter. We follow Gerali et al.
(2010) by setting κb = 10, and κd = 10. These values ensure that the model
approximates the observed interest rate pass-through documented by the
European Central Bank (2009). Gerali et al. (2010) argue that κw is hard to
determine, for this reason they set the standard deviation equal to 5 and de-
cide on a prior value equal to 20 such that the data is able to determine this
parameter. We follow a similar approach. Setting κw = 20 implies that for
each percentage point deviation of the banks capital-to-asset ratio from the
optimal level, the interest rate spread increases by about 4 basis points. We
experimented with different values and found that the data is sufficiently
able to identify this parameter.
As µb and µd are no conventional parameters, no conventional values
are readily available. We therefore calibrate these values to ensure that they
mimic historical averages. In the Appendix Section 3.A we show that in the
steady state µb = rb∗/(rb∗− r∗) and µd = rd∗/(rd∗− r∗). These historical av-
erages are calculated by transforming the quarterly data series representing
annual rates, ry, into quarterly rates, rq, i.e. rq = exp(ln( 1+ry
4 )). Successively,
we use the historical average of these transformed loan, deposit and policy
rates to calculate µd = 4.299 and µb = 2.605. In Gerali et al. (2010) µb is calcu-
lated to be negative which implies that the deposit rate is on average below
the policy rate. Here, we find that the deposit rate is on average higher than
the policy rate. We set the costs of managing bank assets equal to δb = 0.025
close to the value chosen by Gerali et al. (2010). The optimal capital-to-loan
ratio of the bank, ν, is set equal to 8%, the capital requirement imposed by
Basel III on corporate loans.8
8 Basel III prescribes capital requirement based on the risk characteristics of the asset. The
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 78PDF page: 78PDF page: 78PDF page: 78
68 Chapter 3
Following Smets and Wouters (2007) we set the partial indexation para-
meters γp and γw equal to 0.75, the probability of a price and wage update
εp = εw = 0.75 and the wage mark-up λw = 0.5. There is no consensus in
the literature regarding the value of the parameter determining the capital
adjustment costs ϕ. According to Bernanke et al. (1999), reasonable values
lie within the range 0.0− 0.5. Therefore I use a value of 0.25. Furthermore,
δy = 0.75 and δπ = 1.5 which correspond to conventional Taylor rule val-
ues. Finally, we set µ = 0.5 because the value of a credit default swap is
calculated under the assumption of a loss given default of 50%.
The prior specification of the persistence parameters and the variance of
the shocks is shown in Table 3.3. For the value of the persistence parameters
of the shocks ρι, where ι indexes a particular shock, we choose a prior value
of 0.75. As conventional, we choose a beta distribution as prior distribution
for these parameters with a standard deviation equal to 0.1. Finally, the prior
mean and variances of the shock terms are set equal to µι = 0.05 and σι =
0.005 and follow by assumption an inverse gamma distribution.
model simplifies this characteristic and only considers one asset, i.e. loans to firms, whichare homogeneous in risk and have a 100% risk weight.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 79PDF page: 79PDF page: 79PDF page: 79
Credit defaults and bank capital 69
Table 3.2. Table - Estimated parameters
Prior Posterior
Param. Distr. Mean S.D. Mean 10% Median 90%
κw Gamma 20.000 5.000 17.249 12.311 16.826 22.150κd Gamma 10.000 1.000 5.288 4.955 5.227 5.694κb Gamma 10.000 1.000 8.570 7.495 8.532 9.674µb Gamma 4.299 0.400 6.579 6.002 6.620 7.117µd Gamma 2.604 0.200 2.783 2.503 2.776 3.067δy Gamma 0.750 0.050 0.710 0.654 0.709 0.767δπ Gamma 1.500 0.100 1.638 1.514 1.632 1.773γp Gamma 0.750 0.250 0.276 0.190 0.273 0.370γw Gamma 0.750 0.250 0.232 0.151 0.228 0.317µ Gamma 0.500 0.100 0.419 0.326 0.422 0.509ϕ Gamma 0.250 0.100 0.110 0.088 0.110 0.133δb Gamma 0.025 0.010 0.050 0.033 0.049 0.067εp Gamma 0.750 0.050 0.835 0.819 0.836 0.850εw Gamma 0.750 0.050 0.553 0.533 0.552 0.574ν Gamma 0.080 0.005 0.093 0.085 0.092 0.100λw Gamma 0.500 0.050 0.537 0.478 0.537 0.596
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 80PDF page: 80PDF page: 80PDF page: 80
70 Chapter 3
Figure 3.2. Time series plot of observable variables.
00 05 10 15-0.01
-0.005
0
0.005
0.01Policy rate: rt
00 05 10 15-10
-5
0
5#10-3 Loan rate: rb
t
00 05 10 15-2
-1
0
1
2#10-3Deposit rate: rd
t
00 05 10 15-2
0
2
4
6#10-3 Spread: st
00 05 10 15-0.01
-0.005
0
0.005
0.01In.ation: :t
00 05 10 15-0.1
-0.05
0
0.05
0.1Loans to -rms: bt
00 05 10 15-0.04
-0.02
0
0.02
0.04Deposits from households: dt
00 05 10 15-0.05
0
0.05Output: yt
00 05 10 15-0.04
-0.02
0
0.02
0.04Consumption: ct
00 05 10 15-0.1
-0.05
0
0.05
0.1Capital: it
00 05 10 15-0.02
0
0.02
0.04Hours: lt
00 05 10 15-0.04
-0.02
0
0.02
0.04Wages: wt
Notes: Interest rates, inflation and credit spreads are plotted on a quarterly basis and inabsolute deviations from the sample mean. The real variables, loans to firms, deposits tohouseholds, consumption, output, capital and hours worked, are plotted on a quarterly basisand expressed as log deviations from the HP-filtered trend.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 81PDF page: 81PDF page: 81PDF page: 81
Credit defaults and bank capital 71
Tabl
e3.
3.Es
tim
ated
auto
corr
elat
ion
and
stan
dard
devi
atio
npa
ram
eter
s
Para
met
ers
Labe
lD
istr
ibut
ion
Prio
rS.
D.
Mea
n10
%M
edia
n90
%
ρr
Polic
yra
tepe
rsis
tenc
eBe
ta0.
750
0.10
00.
637
0.54
50.
642
0.72
0ρ
aTe
chno
logy
Beta
0.75
00.
100
0.84
30.
798
0.84
70.
884
ρv
Def
ault
Beta
0.75
00.
100
0.64
60.
566
0.64
50.
727
ρrb
Lend
ing
rate
Beta
0.75
00.
100
0.78
00.
670
0.79
10.
870
ρrd
Dep
osit
rate
Beta
0.75
00.
100
0.29
20.
199
0.28
40.
401
ρq
Cap
ital
pric
eBe
ta0.
750
0.10
00.
727
0.68
80.
729
0.76
2ρ
sSp
read
Beta
0.75
00.
100
0.80
30.
695
0.81
20.
894
ρπ
Cos
tpus
hBe
ta0.
750
0.10
00.
327
0.23
90.
324
0.42
1ρ
lLa
bor
supp
lyBe
ta0.
750
0.10
00.
849
0.82
50.
850
0.87
2ρ
iIn
vest
men
tBe
ta0.
750
0.10
00.
664
0.57
10.
672
0.74
5ρ
cC
onsu
mpt
ion
Beta
0.75
00.
100
0.78
50.
742
0.79
00.
822
ρw
Wag
eBe
ta0.
750
0.10
00.
575
0.49
90.
576
0.65
1ρ
rnM
onet
ary
Polic
yBe
ta0.
750
0.10
00.
950
0.93
70.
950
0.96
2σ
aTe
chno
logy
Inv.
gam
0.05
00.
005
0.03
40.
031
0.03
40.
037
σv
Def
ault
Inv.
gam
0.05
00.
005
0.05
20.
045
0.05
10.
060
σ rb
Lend
ing
rate
Inv.
gam
0.05
00.
005
0.04
00.
036
0.04
00.
044
σ rd
Dep
osit
rate
Inv.
gam
0.05
00.
005
0.03
30.
030
0.03
30.
035
σq
Cap
ital
pric
eIn
v.ga
m0.
050
0.00
50.
060
0.05
30.
059
0.06
7σ
sSp
read
Inv.
gam
0.05
00.
005
0.03
80.
034
0.03
80.
041
σπ
Cos
tpus
hIn
v.ga
m0.
050
0.00
50.
042
0.03
80.
042
0.04
6σ
lLa
bor
supp
lyIn
v.ga
m0.
050
0.00
50.
046
0.04
20.
046
0.05
1σ i
Inve
stm
ent
Inv.
gam
0.05
00.
005
0.04
60.
042
0.04
60.
051
σc
Con
sum
ptio
nIn
v.ga
m0.
050
0.00
50.
047
0.04
20.
046
0.05
1σ
wW
age
Inv.
gam
0.05
00.
005
0.04
90.
043
0.04
80.
055
σ rn
Mon
etar
yPo
licy
Inv.
gam
0.05
00.
005
0.03
30.
031
0.03
30.
036
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 82PDF page: 82PDF page: 82PDF page: 82
72 Chapter 3
3.4 Empirical results
Figure 3.3 and 3.4 show the prior and posterior distributions of the estim-
ated parameters and the resulting posterior mode. Table 3.2 shows the cor-
responding summary statistics. The posterior parameter values are drawn
using the Metropolis Hastings algorithm by running 5 chains of 100, 000
draws. The convergence properties of the model are assessed by means of
the convergence statistics suggested by Brooks and Gelman (1998). Over-
all, the convergence statistics indicate that the convergence properties are
satisfied, see also Adjemian et al. (2011) for the appropriate Dynare docu-
mentation.
The posterior distributions suggest that the data is informative about
most estimated parameters. In most cases the posterior distribution is sig-
nificantly different from the prior distribution. However, for λw and δy the
posterior distribution is close to the prior distribution which might indic-
ate that the data is uninformative about these parameters. To determine
whether the data is uninformative, or whether the prior value is simply
close to the value implied by the data, we experimented with different prior
values. Changing the prior mean of λw shifts the posterior distribution. Ap-
parently the data contains no information about this parameter. Fortunately,
the impulse response functions show qualitative similar results for different
values of λw. In contrast, changing δy does not affect the posterior distri-
bution much suggesting that the prior value is relatively close to the value
implied by the data.
3.4.1 Technology shock
The dynamic properties of the model are assessed by comparing the impulse
response functions resulting from a positive technology shock to the im-
pulse response functions generated by canonical DSGE models. Figure 3.5
presents the impulse response functions after a positive technology shock
represented by a one standard deviation increase in the firms’ productivity
level. The parameters are set at the estimated posterior median. In addition
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 83PDF page: 83PDF page: 83PDF page: 83
Credit defaults and bank capital 73
Figure 3.3. Priors and Posteriors real parameters
1.2 1.7 2.20
2
4
/:
0.5 1 1.50
5
.p
0 0.5 1 1.50
5
10
.w
0.2 0.4 0.60
50
'
0.7 0.8 0.90
50
0p
0.6 0.80
10
20
0w
0.6 0.8 10
5
10/y
0.3 0.4 0.5 0.6 0.70
5
6w
Notes: Grey curve: prior; black curve: posterior; dotted line: posterior mode.
we plot the 10% and 90% credibility intervals.
In general the results are comparable to the results presented in the bench-
mark models of Smets and Wouters (2003), Christiano et al. (2005) and Gerali
et al. (2010). The technology shock increases the productivity of both capital
and labor and therefore pushes up the investment schedule and labor de-
mand. Although output increases, inflation decreases because the marginal
cost of production declines. The balance sheet of the entrepreneur extends
because capital income and the real value of physical capital increases in
value. The central bank reacts stronger to inflation than to output and de-
creases the policy rate in response to the occurrence of deflationary pressure.
Banks’ cost of funding determined by the policy rate decline and via the
monetary transmission channel both the lending and deposit rate fall. Con-
sequently, the central bank amplifies the increase in production by respond-
ing stronger to inflation than to output. The inter-temporal substitution ef-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 84PDF page: 84PDF page: 84PDF page: 84
74 Chapter 3
Figure 3.4. Priors and Posteriors financial parameters
20 40 600
0.05
5w
5 100
55b
4 6 8 10 120
0.55d
4 60
1
7b
2.5 3 3.5 40
1
2
7d
0.2 0.6 10
5
7
0.08 0.10
200
8b
0 0.02 0.04 0.060
200
400
/b
Notes: Grey curve: prior; black curve: posterior; dotted line: posterior mode.
fect causes consumption to increase, while the income effect is responsible
for the delayed increase in deposits.
Some notable difference with Gerali et al. (2010) arise due to the specif-
ics of the model presented in this chapter. Via the monetary transmission
channel, both the lending and deposit rate follow the policy rate. The differ-
ence between these two rates determines the banks’ net interest margin. The
technology shock causes the net interest margin to increase and as a result
bank capital increases. Hence, bank capital behaves procyclical which is in
line with empirical findings reported by Albertazzi and Gambacorta (2009).
In Gerali et al. (2010) bank capital is countercyclical because the banks’ in-
terest margin declines. The authors describe this result as a counter-factual
prediction of their model. Whereas the models differ both in structure and
with respect to the estimated parameters, it appears that the fall in the de-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 85PDF page: 85PDF page: 85PDF page: 85
Credit defaults and bank capital 75
fault probability lowers bank provisioning for credit default losses. This ap-
pears sufficient to generate procyclical bank capital dynamics. Moreover,
procyclical bank capital causes the lending rate to fall much stronger than
the deposit rate. As a result, the expansionary effect on credit supply and
investment is amplified.
Figure 3.5. Impulse response functions of a technology shock
0 5 10 15-0.5
00.5
11.5
Output: yt
0 5 10 15-0.5
00.5
11.5
Consumption: ct
0 5 10 15-1
0
1
2Investment: it
0 5 10 15
-2
0
2
Loans: bt
0 5 10 15-2
0
2Deposits: dt
0 5 10 15
-2
0
2
Bank capital: kbt
0 5 10 15-0.2
-0.1
0
0.1Policy rate: rt
0 5 10 15-2
-1
0
1Loan rate: rb
t
0 5 10 15-0.2
-0.1
0
0.1Deposit rate: rd
t
0 5 10 15-0.2
-0.1
0
0.1In.ation: :t
0 5 10 15-0.5
0
0.5Credit spread: st
Notes: Responses to a technology shock se(e Equation (3.12)) represented by the solid lineand the 10% and 90% credibility intervals are presented by the dotted lines. Prices and in-terest rates are shown as absolute deviations from their steady state, expressed in percentagepoints and real variables are percentage deviations from steady state.
3.4.2 Credit default shock
This section describes the effects of a realization of credit defaults different
from anticipated levels which is represented by a one standard deviation in-
crease in unexpected default losses. The parameters are set at the estimated
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 86PDF page: 86PDF page: 86PDF page: 86
76 Chapter 3
posterior median. In addition we plot the 10% and 90% credibility intervals.
Figure 3.6 shows that, following a positive credit default shock, bank
capital falls as the funds reserved for credit default losses are insufficient
to cover the losses. The unanticipated losses deteriorate bank capital and as
a result bank leverage increases. The increase in the lending rate by banks
after a credit default shock is suboptimal from a finance perspective. The de-
cision to finance new projects should be independent from the costs made on
previous projects, i.e., losses on previous projects should be treated as sunk
costs. Credit default losses are sunk costs as these costs will be the same re-
gardless of new lending. However, Figure 3.6 shows that, even though credit
default losses are only a small fraction of the banks’ balance sheet, leverage
forces the banks to lower their lending activities substantially. These results
indicate that credit default losses affect the banks’ ability to issue new debt
and bank equity to finance new credit. Consequently, banks decrease credit
supply via an increase in the lending rate. The real side of the economy ex-
periences an increase in the lending rate and a persistent decline in credit
available for investment. Consequently, the amount of funds available for
investment in the physical capital stock declines.
The credit default shock leads to a small decrease in the inflation rate.
Firms set prices as a mark-up over their marginal costs. While demand con-
tracts and the wage rate falls, firms borrow to finance production and the
cost of borrowing increases. The net effect is a marginal decrease in the in-
flation rate on impact. The marginal decrease of inflation is consistent with
theoretical evidence presented by Christiano et al. (2005) and Gerali et al.
(2010). Moreover, Gilchrist et al. (2017) present empirical evidence for the
presence of a “cost channel” in the U.S.: when firms have weak balance
sheets, they may pass on costs increases, i.e., higher lending rates, to their
customers in the form of higher prices. These results suggest that a credit
default shock does not have much impact on inflation.
Notably, as banks are constrained by their leverage ratio, monetary trans-
mission is impeded and lending rates remain high despite monetary accom-
modation. The central bank follows a Taylor rule and lowers the policy rate
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 87PDF page: 87PDF page: 87PDF page: 87
Credit defaults and bank capital 77
Figure 3.6. Impulse response functions of a credit default shock
0 5 10 15-0.5
0
0.5Output: yt
0 5 10 15-0.2
0
0.2Consumption: ct
0 5 10 15-1
0
1Investment: it
0 5 10 15-4
-2
0Loans: bt
0 5 10 15-2
-1
0Deposits: dt
0 5 10 15-4
-2
0Bank capital: kb
t
0 5 10 15-0.2
0
0.2Policy rate: rt
0 5 10 15-0.5
0
0.5Loan rate: rb
t
0 5 10 15-0.2
0
0.2Deposit rate: rd
t
0 5 10 15-0.05
0
0.05In.ation: :t
0 5 10 15-0.5
0
0.5Credit spread: st
Notes: Responses to a credit default shock (see Equation (3.18)) represented by the solid lineand the 10% and 90% credibility intervals are presented by the dotted lines. Prices and in-terest rates are shown as absolute deviations from their steady state, expressed in percentagepoints and real variables are percentage deviations from steady state.
because both output and inflation fall. While the increase in bank leverage
affects the lending rate directly, it does not affect the deposit rate which
closely follows the policy rate. As a result the lending rate increases while
the deposit rate falls. The decrease in the deposit rate reduces savings and
increases consumption of patient households. Entrepreneurs, however, de-
crease consumption because the lending rate has increased. Hence, monet-
ary policy only affects the recovery of household consumption while invest-
ment and entrepreneurial consumption remain subdued despite expansion-
ary monetary policy.
In conclusion, when banks are undercapitalized and cannot or will not
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 88PDF page: 88PDF page: 88PDF page: 88
78 Chapter 3
issue new bank equity, monetary transmission is impeded. In this case, cut-
ting the policy rate resembles pushing a string as the banks’ cost of funds is
not the binding constraint that restricts credit supply. Decreasing the policy
rate has little direct impact. Indirectly, however, the central bank recapital-
izes the banks. The fall in the policy rate allows banks to lower their deposit
rate while the lending rate remains high. When the bank capital constraint
does not bind, arbitrage ensures that the lending rate falls in accordance
with the policy rate. However, as the entire banking sector is constrained
by their leverage ratio, the interest rate spread is not arbitraged away. Con-
sequently, the banks’ profit margin increases and banks rebuild their bank
capital from retained earnings. Lowering the policy rate thus effectively sub-
sidizes banks by taxing savers and realizes a relatively slow bank recapital-
ization.
Historical decomposition
Figure 3.7 and 3.8 show the historical decompositions of output, inflation,
consumption and investment. The historical decompositions identify a se-
quence of shocks that are able to account for the dynamics of the endogen-
ous variables and thereby provide insight in the economic impact of partic-
ular shocks. The decomposition of output shows that credit default shocks
had considerable impact on output fluctuations. In particular, during the
run-up to the global financial crisis, lower than anticipated realizations of
credit defaults drove output above potential while after the burst credit
default shocks had a noticeable adverse effect on output. The occurrence
of these credit default shocks appears largely in synchronization with the
boom-bust cycle suggesting that credit default losses have been an import-
ant determinant of historical fluctuations in output.
The historical decomposition of inflation suggests, consistent with the
interpretation of the impulse response functions above, that credit default
shocks only had minor impact on inflation dynamics. In accordance with
the pre-crisis consensus view of inflation drivers, inflation dynamics appear
mostly driven by monetary policy shocks. Another driver, although less fre-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 89PDF page: 89PDF page: 89PDF page: 89
Credit defaults and bank capital 79
Figure 3.7. Historical decomposition output and inflation
Notes: Horizontal axis shows years. Vertical axis shows deviations from steady state. Thevariables on the right hand side denote respectively: wage shock εw
t , investment shock εit,
labor supply shock, εlt, credit spread shock εs
t, capital price shock εlt, preference shock εc
t ,credit default shock εv
t , deposit demand shock εrdt , lending demand shock εrb
t , inflation shockεπ
t , technology shock εat , monetary policy shock εrn
t ,
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 90PDF page: 90PDF page: 90PDF page: 90
80 Chapter 3
Figure 3.8. Historical decomposition consumption and investment
Notes: Horizontal axis shows years. Vertical axis shows deviations from steady state. Thevariables on the right hand side denote respectively: wage shock εw
t , investment shock εit,
labor supply shock, εlt, credit spread shock εs
t, capital price shock εlt, preference shock εc
t ,credit default shock εv
t , deposit demand shock εrdt , lending demand shock εrb
t , inflation shockεπ
t , technology shock εat , monetary policy shock εrn
t ,
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 91PDF page: 91PDF page: 91PDF page: 91
Credit defaults and bank capital 81
quently, is the technology shock. Technology and monetary policy shocks
often cancel out indicating a relative aggressive policy response to techno-
logy shocks compared to what the conservative Taylor rule in the model
predicts. The impulse response functions in Figure 3.5 indeed suggest that
if the central bank follows a conventional Taylor rule, it does not respond
aggressively to a technology shock.
Decomposing output fluctuations in consumption and investment fluc-
tuations offers a more comprehensive assessment of the credit default shock
transmission channel. Credit default shocks appear to have had consider-
able impact on historical fluctuations in investment. In contrast, consump-
tion seems mostly unaffected by credit default shocks. These results indicate
that especially the credit supply channel of credit default shocks is relevant
for output fluctuations and supports the notion that an undercapitalized
banking sector suppresses investment.
3.4.3 Countercyclical capital buffer
When banks are undercapitalized and cannot issue new bank equity, mon-
etary transmission is impeded. Lowering the policy rate has little direct im-
pact on investment because the lending rate remains high. For this reason,
we introduce a countercyclical capital buffer to alleviate the bank capital
constraint directly. Figure 3.9 shows the response to a default shock when
the central bank implements the countercyclical capital buffer which is ac-
tivated when the credit-to-ouptut ratio differs from its steady state ratio spe-
cified by Equation (3.31). As argued before, we also assume that the coun-
tercyclical capital buffer is credible in the sense that the market does not
require a higher level of bank capital, i.e., in any case only the regulatory
capital requirement is binding.
We experimented with different values for the policy reaction parameter
$y. The solid line represents the baseline model and sets $y = 0 (no coun-
tercyclical capital buffer), the dashed line sets $y = 1 and the dotted line
sets $y = 2 while keeping all other parameters the same. The calibration res-
ults in Figure 3.9 show that the countercyclical capital buffer attenuates the
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 92PDF page: 92PDF page: 92PDF page: 92
82 Chapter 3
Figure 3.9. Impulse response functions of a credit default shock and a coun-tercyclical capital buffer based on credit-to-output.
5 10 15
-0.2
-0.1
0
Output yt
5 10 15
-1.5
-1
-0.5
0Bank capital: kb
t
5 10 15
-20
-10
0
#10!3 In.ation: :t
5 10 150
0.2
0.4
0.6
0.8
Bank leverage: bt
kbt
0.15
%y = 0
%y = 1
%y = 2
Notes: prices and interest rates are shown as absolute deviations from steady state, expressedin percentage points and real variables are percentage deviations from steady state. The solidline represents the benchmark model and sets $y = 0 (no countercyclical capital buffer), thebar stripes (−−−) represents $y = 1 and the dotted line (· · · ) represents $y = 2.
fluctuations in output and inflation caused by a credit default shock. In gen-
eral, a larger response of the countercyclical buffer relative to the change in
the credit-to-output ratio attenuates macroeconomic fluctuations. Nonethe-
less, the marginal benefit is declining as can be seen by comparing the one-
for-one response ($y = 1) and the two-for-one response ($y = 2) with the
baseline model. The intuition behind this result is as follows. When credit
is used solely for investment without affecting consumption and labor sup-
ply, credit and output move one-for-one and their ratio is unaffected. How-
ever, when the bank capital constraint is relaxed by activating the counter-
cyclical capital buffer, credit supply increases by more than output because
also debt-financed consumption becomes higher. As a consequence, the ini-
tial change in the credit-to-output ratio is reduced and the marginal effect of
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 93PDF page: 93PDF page: 93PDF page: 93
Credit defaults and bank capital 83
Figure 3.10. Impulse response functions of a credit default shock and a coun-tercyclical capital buffer based on steady state bank capital.
5 10 15
-0.2
-0.1
0
Output yt
5 10 15
-1.5
-1
-0.5
0Bank capital: kb
t
5 10 15
-20
-10
0
#10!3 In.ation: :t
5 10 150
0.5
1
Bank leverage: bt
kbt
%y = 0
%y = 0:5
%y = 0:9
Notes: prices and interest rates are shown as absolute deviations from steady state, expressedin percentage points and real variables are percentage deviations from steady state. The solidline represents the benchmark model and sets $k = 0 (no countercyclical capital buffer), thebar striped line (−−−) represents $k = 0.5 and the dotted line (· · · ) represents $k = 0.9.
the countercyclical buffer diminishes.
In reality, credit and output are potentially less correlated than in the
model. For example, credit here also accounts for mortgage loans and con-
sumer credit, while one could argue that output does not benefit directly
from these forms of credit.9 Consequently, we might observe only small
changes in the credit-to-output ratio, while in fact financial imbalances build-
up quite rapidly. It is therefore more advantageous to specify a countercyc-
lical capital buffer that is contingent on bank capital. Figure 3.10 shows the
responses to a default shock when the central bank activates the counter-
9 See for example Bernanke (2005), Mian and Sufi (2014) and Mian et al. (2016) who arguethat an increase in the household debt-to-output ratio leads to lower future output growth.In Chapters 4 and 5 we discuss these issues more extensively.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 94PDF page: 94PDF page: 94PDF page: 94
84 Chapter 3
cyclical capital buffer when bank capital differs from steady state bank cap-
ital specified by (3.32). The solid line represents the baseline model and sets
$k = 0, the bar striped line represents $k = 0.5 and the dotted line represents
$k = 0.9.
The results show that output and inflation fluctuations are further re-
duced compared to the countercyclical buffer contingent on credit-to-output.
These results are somewhat trivial as $k = 1 dissolves any feedback between
bank leverage and the lending rate and restores monetary transmission.
However, the results illustrate that countercyclical capital buffers increase
the persistence of the downturn because the deterioration of bank capital
last longer. Whereas the increase in bank leverage is precisely what the coun-
tercyclical buffer enables, it also reduces incentives to rebuild bank capital
as it reduces the costs of deviating from the capital requirement. The in-
crease in persistence is only small for output and inflation, but significant
for bank capital. Recently, Shin (2014) showed that since the global financial
crisis banks have been very slow in rebuilding bank capital and preferred
dividend payouts over the retention of earnings. The results in this chapter
indicate that activating countercyclical capital buffers in practice might fur-
ther lower bank incentives to rebuild bank capital or issue new equity. The
activation of these buffers can therefore adversely impact financial stability
in the long-run.
3.4.4 Endogenous recapitalization
A potential solution to overcome the reduced incentives to rebuild or is-
sue new bank equity is a mandatory bank recapitalization. Credit default
losses are sunk costs, but the estimation results show that credit default
shocks do affect investment. The countercyclical capital buffer analyzed in
the previous section attenuates fluctuations as it allows banks to operate at
a lower leverage ratio. However, if activated, banks take even more time
to rebuild their bank equity without accumulating any of the insolvency
costs. An endogenous recapitalization could mitigate this incentive prob-
lem as rebuilding bank equity is no longer the bank’s choice. Recapitaliza-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 95PDF page: 95PDF page: 95PDF page: 95
Credit defaults and bank capital 85
Figure 3.11. Impulse response functions of a default shock and a counter-cyclical capital buffer based on steady state bank leverage.
5 10 15
-0.2
-0.1
0
Output yt
5 10 15
-1.5
-1
-0.5
0Bank capital: kb
t
5 10 15
-20
-10
0
#10!3 In.ation: :t
5 10 15
-5
0
5
10
15
#10!3Bank leverage: bt
kbt
%y = 0
%y = 0:25
%y = 0:50
Notes: prices and interest rates are shown as absolute deviations from steady state, expressedin percentage points and real variables are percentage deviations from steady state. The solidline represents the baseline model and sets $kb = 0 (no countercyclical capital buffer), the barstriped line (−−−) represents $kb = 0.25 and the dotted line (· · · ) represents $kb = 0.5.
tions are effective because household deposits are effectively converted into
bank equity without changing the claim of households on the cash flows of
banks. Consequently, if the irrelevance proposition of Modigliani and Miller
(1958) holds, households and banks are unaffected by the recapitalization.
For this reason, as discussed in Section 3.2.3, we assume that households
prefer deposits over bank equity because deposits are more liquid. This as-
sumption causes a violation of the Modigliani-Miller irrelevance proposi-
tion as in Stein (2012).
As argued, the bank recapitalization can be specified in multiple ways.
Figure 3.11 shows the effects of an endogenous bank recapitalization which
is financed by the government granting the bank an interest free loan. The
recapitalization is specified by exactly the same processes as the countercyc-
lical capital buffers in the previous section. The results show that an endo-
genous recapitalization is very effective in attenuating the impact of a credit
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 96PDF page: 96PDF page: 96PDF page: 96
86 Chapter 3
default shock on the real economy. Both the amplification as well as the
persistence in output and inflation fluctuations after a credit default shock
decrease substantially. These results suggest that taxing the consumers in
terms of foregone interest rate savings has only limited impact on output
and inflation. Moreover, as bank capital is supplemented when it deteri-
orates, also bank leverage remains low which increases the stability of the
financial system.
3.5 Conclusion
In this chapter we analyzed the effects of credit default losses on the bank-
ing sector and the real economy by extending a standard macro-model with
a banking sector and credit default risk. We fitted the model to euro area
data. The results show that when credit default losses are at a higher level
than anticipated—a credit default shock—bank capital deteriorates causing
bank leverage to rise and credit supply to fall. These results indicate that
banks do not issue new bank equity after a credit default shock to rebuild
their bank equity. Consequently, credit supply is constrained because banks
have a high leverage ratio. Also monetary policy is impeded because lend-
ing rates remain high despite expansionary monetary policy. Output falls
while inflation is less affected because firms set prices as a mark-up over
their marginal costs and funding costs increase. The historical decomposi-
tions show that credit default shocks are an important driver of historical
output fluctuations.
The model is able to explain the slow recovery of economic activity after
the global financial crisis despite unprecedented expansionary monetary
policy. As long as banks are undercapitalized, monetary transmission is im-
peded. Conventional monetary policy cannot invigorate credit supply in
this case because the lending rate remains high. Central banks do not lend
directly to firms and can therefore not circumvent the monetary transmis-
sion channel. The historical decompositions do indeed show that credit de-
fault shocks have been an important driver of output and investment fluctu-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 97PDF page: 97PDF page: 97PDF page: 97
Credit defaults and bank capital 87
ations during the recent global financial crisis. Comparable to a classic debt
overhang interpretation, credit supply is restricted because losses on exist-
ing credit restrict the issuance of new debt and new bank equity for new
credit.
Expansionary monetary policy is accommodating as it helps the banks
to recapitalize by increasing the banks’ net interest margin. However, accu-
mulating bank capital from retained earnings is a slow processes. For this
reason, we allowed for two alternative policy instruments that both address
the bank capital constraint directly. Both instruments, the countercyclical
capital buffer and the recapitalization, complement conventional monet-
ary policy and support the macroeconomic recovery. It is unclear, however,
whether both measures will be as effective in practice as they are in theory.
For one thing, we ignored the presence of moral hazard. Moreover, relaxing
leverage constraints is meaningless when the market leverage constraint be-
comes binding instead. We leave these issues for future research.
3.A Model solution
Household maximization problem
Households maximize their utility subject to their budget constraints and
labor demand. The household Lagrangian is represented by:
Lht ≡ Et
∞
∑t=0
(βH)t
ηc
t
([CH
t (i)− hCHt−1(i)
]1−σc
1− σc−
ηlt[LH
t (i)]1+σh
1 + σh
)+
λh,1t (i)
(wt(i)LH
t (i) +1 + rd
t−1
πtDt−1(i) + RPt(i)− CH
t (i)− Dt(i)
)+
λh,2t (i)
LHt (i)−
(wt(i)
wt
)− 1+λwt
λwt LH
t
, (3.A.1)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 98PDF page: 98PDF page: 98PDF page: 98
88 Chapter 3
where λh,1t (i) and λh,2
t (i) are the Lagrangian multipliers w.r.t. the budget
constraint and labor demand. The FOCs w.r.t. CHt (i), Dt(i) and LH,d
t are
given by:
ηct (C
Ht (i)− hCH
t−1(i))−σc = λh,1
t (i), (3.A.2)
Et
λh,1
t+1(i)1 + ρH
1 + rdt
πt+1
= λh,1
t (i), (3.A.3)
ηct ηl
t
[LH
t (i)]σh
= λh,2t (i)wt + λh,2
t (i). (3.A.4)
The household is represented by a labor union which can only change the
wage and sets it optimally at a level wt with a probability ξw. Otherwise, it
uses the wage indexation rule (3.4). Hence, the wage rate equals:
wt(i) =
wt(Pt−1Pt−2
)γw
wt−1(i)
if set optimally,
otherwise,(3.A.5)
To derive the household’s FOCs with respect to the wage rate in those mar-
kets where the wage rate is set optimally in the current period, we reproduce
the part of the Lagrangian that is relevant:
Lht = Et
∞
∑s=0
ηct+s(βjξw)s
−
ηlt+s[LH
t+s(i)]1+σh
1 + σh+ λt+s(i)wt+s(i)LH
t+s(i)
,
(3.A.6)
subject to labor demand LHt (i) =
(wt(i)
wt
)− 1+λwt
λwt LH
t . Substituting labor de-
mand and using εw ≡ 1+λwt
λwt
gives:
Lht = Et
∞
∑s=0
(βjξw)s
−
ηlt+s
1 + σh
[(wt+s(i)
wt+s
)−εw
LH,dt+s
]1+σh
+
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 99PDF page: 99PDF page: 99PDF page: 99
Credit defaults and bank capital 89
λt+s(i)(
Pt
Pt−1
)γw(
w1−εwt+s (i)wεw
t+s
)LH,d
t+s
. (3.A.7)
Taking the derivative w.r.t. wt gives
w1+εwσht (i)
∞
∑s=0
ηct+s(βjξw)s LH,d
t+s(CHt+s(i)− hCH
t+s−1(i))−σc
(1 + λwt )
=
Et
∞
∑s=0
(βjξw)s(
ηlt+s
[LH,d
t+s
]σhwεwσh
t+s
), (3.A.8)
and using that s periods after the optimization wt+s =(
Pt+sPt+s−1
)γwwt+s−1 and
rewriting (3.A.8) gives:
wt(i)Pt
Et
∞
∑s=0
βs(ξw)s
(
PtPt−1
)γw
Pt+sPt+s−1
LHt+s(C
Ht+s(i)− hCH
t+s−1(i))−σc
1 + λw−
(ηl
t+s
[LH,d
t+s
]σh)
= 0.
(3.A.9)
Using the wage indexation rule, it is possible to write the aggregate wage
process as:
W− 1
λwt = ξw
((Pt−1
Pt−2
)γw
Wt−1
)− 1λw
+ (1− ξw)w− 1
λwt . (3.A.10)
Entrepreneur maximization problem
Entrepreneurs maximize their utility subject to the budget constraint, the
participation constraint of the banks and the capital accumulation identity:
Let ≡Et
∞
∑t=0
ηct (βE)t
ln
CEt − hCE
t−1
+
λet
([1− Γ(ωt+1)]
[rk
t+1ut − ψ(ut)]
Kt + Bt−
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 100PDF page: 100PDF page: 100PDF page: 100
90 Chapter 3
1 + rbt−1[1− Γ(ωt)]
πtBt−1 − It − wtLH
t − CEt
)+
λet φt
([Γ(ωt+1)− µG(ωt+1)]rk
t+1utqtKt −rwb
tπt+1
Bt
)+
λetqt
((1− δk)Kt−1 +
[1− ψ
(ηi
t It
It−1
)]It − Kt
)(3.A.11)
where wt =(
wHt
Ω
)Ω ( wEt
1−Ω
)1−Ω. The FOCs w.r.t. CE
t , Kt, Bt, ωt+1, It and ut,
using the bank participation constraint with equality denote:10
ηct
CEt − hCE
t−1= λe
t , (3.A.12)
([1− Γ(ωt+1)](rk
t+1ut − ψ(ut))− qt
)+
λet+1
λet
βEqt+1(1− δk)+
φt
([Γ(ωt+1)− µG(ωt+1)]rk
t+1utqt
)= 0, (3.A.13)
λet+1βE 1 + rb
t [1− Γ(ωt+1)]
πt+1+ φtλ
et
rwbt
πt+1= λe
t , (3.A.14)
Γ′(ωt+1)(rk
t+1ut − ψ(ut)) =φt[Γ′(ωt+1)− µG
′(ωt+1)]rk
t+1utqt+
λet+1
λet
rbt Γ′(ωt+1)
πt+1
Bt
Kt(3.A.15)
λet qt
([1− ψ
(ηi
t It
It−1
)− ψ
′(
ηit It
It−1
)ηi
t It
It−1
]− 1)+
λet+1qt+1ψ
′(
ηit It
It−1
)(ηi
t It
It−1
)(It
It−1
)= 0, (3.A.16)
10 We assume free entry and exit in the banking sector such that competition equalizes thelending rate to the weighted average cost of funding.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 101PDF page: 101PDF page: 101PDF page: 101
Credit defaults and bank capital 91
[1− Γ(ωt+1)](rkt+1 − ψ
′(ut))Kt + φt[Γ(ωt+1)− µG(ωt+1)]rk
t+1qtKt = 0.
(3.A.17)
and the three constraints stated in (3.A.11).
Firms maximization problem
Retailers buy the intermediate products produced in the intermediate goods
sector and transform it into a homogeneous good Yt using a CES production
function, see Dixit and Stiglitz (1977):
Yt =
[∫ 1
0Yt(j)1−1/λp di
]1/(1−1/λp)
, (3.A.18)
where Yt(j) is an unique input variety produced by entrepreneur j and λp
is the elasticity of substitution in production. The retailer minimizes costs∫ ∞0 Pt(j)Yt(j) subject to the CES production function, Equation (3.A.18). Hence,
the retailer optimization problem becomes:
Lt ≡∫ 1
0Pt(j)Yt(j)di + λt
[Yt −
[∫ 1
0Yt(j)1−1/λp di
]1/(1−1/λp)]
. (3.A.19)
The solution defines unit costs Pt and demand for Yt(j):
Pt ≡[∫ 1
0P1−λp
t di]1/1−λp
, (3.A.20)
Yt(j) =Yt
[Pt(j)
Pt
]−λp
. (3.A.21)
Entrepreneurs own firms therefore the maximization problem of firm i is
part of the entrepreneurs decision problem. From Equation (3.12) we derive
the intermediate firm optimization problem. This is an intermediate step to
determine the optimal capital-labor mix:
minKt(i),LH
t (i)rk
t Kt(i) + WtLHt (i),
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 102PDF page: 102PDF page: 102PDF page: 102
92 Chapter 3
subject to Yt(i) = At[Kt(i)αLHt (i)
1−α]. (3.A.22)
Cost minimization gives the following optimal capital-labor condition:
WtLHt (i)
rkt Kt(i)
=1− α
α, (3.A.23)
which implies that the capital-labor ratio is identical across firms. Firms’
marginal costs are given by:
MCt =1At
(rk
tα
)α( Wt
1− α
)1−α
, (3.A.24)
where MCt denotes the intermediate firms’ marginal costs. Firms maximize
expected firm value by setting prices Pt(i):
Πt(i) ≡NPt(i)
Pt=
[Pt(i)
Pt− MCt
Pt
]Yt
(Pt(i)
Pt
)−λpt
. (3.A.25)
Firms maximize their real profits Πt(i) by choosing the price level Pt(i). Us-
ing Calvo Pricing (Calvo, 1983), and denoting the probability that a firm is
able to change its price by the probability (1− εp), we obtain the expected
value of the firm that has just received a “green light”, i.e., the firm is al-
lowed to change its price in period t, and has set a new price Pnt (i):
maxPn
t (i)Et
∞
∑s=0
(εp)sβE[
Pnt (i)
Pt+s− MCt+s
Pt+s
]Yt+s
(Pn
t (i)Pt+s
)−λpt
. (3.A.26)
Maximizing Equation (3.A.26) w.r.t. the new price Pnt (i) and rewriting gives:
Pnt (i) = Pn
t =λ
pt
λpt − 1
Et∑∞
s=0(εp)sβEPλ
pt
t+sYt+sMCt+s
Pt+s
∑∞s=0(ε
p)sβEPλpt −1
t+s Yt+s
. (3.A.27)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 103PDF page: 103PDF page: 103PDF page: 103
Credit defaults and bank capital 93
As standard, the law of motion of the price level is given by:
(Pt)1−λ
pt = ξ p
((Pt−1
Pt−2
)γp
Pt−1(i))1−λ
pt
+ (1− ξ p)(Pnt )
1−λpt . (3.A.28)
Finally, firms do not incorporate the marginal effect they have on the
aggregate probability of default of all firms. Besides, households perfectly
diversify between all the firms in the economy. The law of large numbers
ensures that the representative households receives real dividend payments
equal to:
Πt = [1− F(ω)]
[1− MCt
Pt
]Yt. (3.A.29)
Banks
Loan and deposit demand
The demand function for loans is derived in a similar fashion as demand for
product Yt(i). Entrepreneurs minimize their loan payments subject to the
aggregation technology, see Dixit and Stiglitz (1977):
Lbt ≡
∫ 1
0rb
t (j)Bt(j)dj + λbt
[Bt −
[∫ 1
0Bt(j)1−1/µb dj
]1/(1−1/µb)]
. (3.A.30)
Solving this problem by aggregating all FOCs across all entrepreneurs gives
loan demand at bank j:
Bt(j) =(
rbt (j)rb
t
)µb
Bt. (3.A.31)
In a similar fashion the deposit demand function is derived. Households
maximize interest revenues from deposits subject to the aggregation tech-
nology:
Ldt ≡
∫ 1
0rd
t (j)Dt(j)dj + λdt
[Bt −
[∫ 1
0Dt(j)1−1/µd dj
]1/(1−1/µd)]
. (3.A.32)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 104PDF page: 104PDF page: 104PDF page: 104
94 Chapter 3
Solving this problem by aggregating all FOCs across all household gives
deposit demand at bank j:
Dt(j) =(
rdt (j)rd
t
)µd
Dt. (3.A.33)
Retail Branch
The retail branch of bank j consists out of 2 parts, a lending branch and a
funding branch. Both maximize their profits subject to the demand sched-
ules by choosing the appropriate interest rates. Substituting demand for
loans, Equation (3.16), in Equation (3.25) gives the following maximization
problem:
maxrb
t (j)
∞
∑t=0
λh,1t
(rbt (i)− rwb
t )
(rb
t (j)rb
t
)µb
Bt −κb
2
(rb
t (j)rb
t−1(j)− 1
)2
rbt Bt
,
(3.A.34)
The solution to the problem is:
(rb
t (j)rb
t
)µb
Bt
((1 + µb)− µb rwb
t
rbt (j)
)+ κb
(1
rbt−1(j)
− rbt (j)
(rbt−1(j))2
)rb
t Bt+
κbβHt
λh,1t+1
λ1,ht
Et
(rb
t+1(j)rb
t (j)− 1
)(rb
t+1(j)rb
t (j)
)2
Bt+1
= 0.
(3.A.35)
The solution to the lending branch optimization problem, dropping the bank
indexation parameter ` which imposes a symmetric equilibrium (rbt (j) ≡ rb
t ,
∀j), yields:
−κb
[(1− rb
t
rbt−1
)rb
t
rbt−1
+βHt
λh,1t+1
λ1,ht
Et
(rb
t+1
rbt− 1)(
rbt+1
rbt
)2 Bt+1
Bt
]=
1− µb + µb rwbt
rbt
, (3.A.36)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 105PDF page: 105PDF page: 105PDF page: 105
Credit defaults and bank capital 95
where λh,1t denotes the multiplier on the patient household budget con-
straint (3.3). Notice that if Equation (3.A.36) is log-linearized we obtain Equa-
tion (3.A.56).
In a similar fashion the funding branch maximizes its profits subject to
the deposit demand schedule. Substituting demand for deposits, Equation
(3.17), in Equation (3.26) gives the following maximization problem:
maxrd
t (j)E0
∞
∑t=0
λh,1t
(rt − rdt (j))
(rd
t (j)rd
t
)µd
Dt −κd
2
(rd
t (j)rd
t−1(j)− 1
)2
rdt Dt
.
(3.A.37)
The solution, after rewriting in a similar fashion as in Equation (3.A.35), is:
(rd
t (j)rd
t
)µd
Dt
(− 1 + µd − µd rt
rdt (j)
)+ κd
[(1
rdt−1(j)
− rdt (j)
(rdt−1(j))2
)rd
t Dt
+ βHt
λh,1t+1
λ1,ht
Et
(rd
t+1(j)rd
t (j)− 1
)(rd
t+1(j)rd
t (j)
)2
Dt+1
]= 0. (3.A.38)
The solution to the funding branch optimization problem, dropping the
bank indexation parameter ` which imposes a symmetric equilibrium (rdt (j)
≡ rdt , ∀j), yields:
κd
[(1− rd
t
rdt−1
)rd
t
rdt−1
+ βHt
λh,1t+1
λ1,ht
Et
(rd
t+1
rdt− 1)(
rdt+1
rdt
)2 Dt+1
Dt
]=
1− µd + µd rt
rdt
. (3.A.39)
The log-linear model
For the empirical analysis we linearize the model around the non-stochastic
steady state. A hat denotes a log-linearized variable. The household con-
sumption Euler is obtained by combining (3.A.3) and (3.A.2) in linearized
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 106PDF page: 106PDF page: 106PDF page: 106
96 Chapter 3
form represented by:
cHt =
h1 + h
cHt−1 +
11 + h
EtcHt+1 −
1− h1 + h
(rd
t − Etπt+1)+
1− h1 + h
(ηct − Etηc
t+1) . (3.A.40)
This is the conventional forward-looking consumption equation with ex-
ternal habit formation, i.e., consumption today depends on past consump-
tion and expected consumption.
Linearizing the wage setting equation, (3.A.9) and (3.A.10), gives:
wt =βH
1 + βH
(Etwt+1+ Etπt+1+
1βH wt−1
)− 1 + βHγw
1 + βH πt+
γw
1 + βH πt−1 −1
1 + βH(1− βHεw)(1− εw)(
1 + (1+λw)σlλw
)εw
×
(wt − σl lH
t −1
1− h(cE
t − cEt−1)− ηl
t
). (3.A.41)
The wage rate depends via partial indexation positively on the past wage
rate and via Calvo pricing positively on the expected future wage rate .
The entrepreneur consumption Euler is given by combining (3.A.12) and
(3.A.14). Linearizing both equations yields:
− 11− h
cEt +
h1− h
cEt−1 + ηc
t = λet , (3.A.42)
and (λe
t+1 + rbt − πt+1
)−(
1[1− Γ(ω)]
)Γ′(ω)ω ˆωt+1+(
φrb
1− φrb
)(φt + λe
t + rwbt − πt+1
)= λe
t , (3.A.43)
where
rb = (1− βE)βE[1− Γ(ω)] + φ (3.A.44)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 107PDF page: 107PDF page: 107PDF page: 107
Credit defaults and bank capital 97
and
φ =
(Γ′(ω)
[Γ′(ω)− µG′(ω)]
)(1− rb
rkBK
)(3.A.45)
where ω = rbB/rkK. Consumption of the entrepreneur evolves in an sim-
ilar fashion as consumption of the households; it depends on past and on
expected consumption. However, entrepreneurial consumption depends, in
contrast to households consumption, on the real lending rate rather than
the real deposit rate. Moreover, the probability of default also affects the
consumption decision of the entrepreneur.
The investment equation (3.A.16) is standard and in linearized form rep-
resented by:
it =1
1 + βE it−1 +βE
1 + βE Etit+1+1
1 + βE1ϕ
qt +1
1 + βE ηit−
βE
1 + βE Etηit+1, (3.A.46)
where ϕ = ψ′′. Investment depends on past and expected investment via
the capital adjustment costs function.
The capital pricing equation (3.A.13) is linearized and represented by:
qt =[1− Γ(ω)]rk rkt+1 + [1− Γ(ω)]
(rk − ψ
′(u))
ut+
βE(1− δk)(λe
t+1 − λet + qt+1
)+
[Γ(ω)− µG(ω)]φrk(
φt + rkt+1 + ut + qt
), (3.A.47)
where
rk =1− βE (1− δk)
[1− Γ(ω)] + φ[Γ(ω)− µG(ω)](3.A.48)
which is higher than the steady state return to capital in Smets and Wouters
(2003) because firms need to pay banks a mark-up for the possibility of bank-
ruptcy.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 108PDF page: 108PDF page: 108PDF page: 108
98 Chapter 3
Linearizing the FOC for the default threshold (3.A.15) gives:
Γ′′(ω)(rk − ψ(1))ω ˆωt+1 + Γ
′(ω)rk
(rk
t+1 + ut
)=
[Γ′(ω)− µG
′(ω)]rkφ
(φt + rk
t+1 + ut + qt
)+
[Γ′′(ω)− µG
′′(ω)]rkφω ˆωt+1+
rbΓ′(ω)
b∗
k∗(
λet+1 − λe
t+1 + rbt − πt+1 + bt − kt
)+ rbΓ
′′(ω)
b∗ωk∗
ˆωt+1,
(3.A.49)
where lower case letters with a star (∗) denote steady state values. Moreover,
(3.A.49) argues that both a lower return to capital, a lower payback prob-
ability of the loan and the bank participation constraint affect the defaults
threshold. Linearizing the FOC for the utilization rate (3.A.17) gives:
[1− Γ(ω)][rk
t rkt+1 + ψ
′′(u)ut
]− Γ
′(ω)(rk − ψ
′(u))ω ˆωt+1+
[Γ(ω)− µG(ω)]rkφ(
φt + rkt+1 + qt
)+ rkΓ
′(ω)ω ˆωt+1 = 0, (3.A.50)
which is different from Smets and Wouters (2003) because in this model the
probability of default interacts with the capital utilization rate. Linearizing
the bank participation threshold (3.11) gives:
[Γ′(ω)− µG
′(ω)]
[Γ(ω)− µG(ω)]ω ˆωt+1 =
(rwb
t − πt+1 + bt
)−(
rkt+1 + qt + kt + ut
),
(3.A.51)
which states that the default threshold is at the aggregate level determined
by aggregate entrepreneurial leverage.
The model of the production side of the economy is standard. The Cobb-
Douglas production function is represented by:
y = at + αkt−1 + αψrkt + (1− α)lt. (3.A.52)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 109PDF page: 109PDF page: 109PDF page: 109
Credit defaults and bank capital 99
The optimal capital labor condition is given by:
lHt = −wt + rk
t + kt−1, (3.A.53)
and the inflation rate is given by:
πt =βE
1 + βEγpEtπt+1+
γp
1 + βEγpπt−1+
11 + βEγp
(1− βEεp)(1− εp)
εp(αrk
t + (1− α)wt − ηat + η
pt ).
(3.A.54)
The financial side of the economy is represented by the lending rate set
by the bank holding company:
rwbt − rt =−
κwν3
r(kb
t − bt). (3.A.55)
Hence, the spread between the lending rate set by the bank holding com-
pany and the risk-free rate (rwbt − rt) can be either positive or negative de-
pending on leverage. If, for example, leverage is low such that the capital-
to-asset ratio is above the optimal capital-to-asset ratio ν, bank capital kbt is
sufficient to cover the outstanding loans bt. The interest spread will have to
rise to decrease the amount of outstanding loans.
Log-linearizing the loan rate setting equation gives:
rbt = ζb
1rbt−1 + ζb
2Etrbt+1+ ζb
3rwbt , (3.A.56)
where ζb1 ≡ κb/(µb − 1 + (1 + βH)κb), ζb
2 ≡ βHκb/(µb − 1 + (1 + βH)κb),
ζb3 ≡ (µb − 1)/(µb − 1 + (1 + βH)κb). Equation (3.A.56) states that the loan
rate depends on the loan rate in the previous period, the expected loan rate
in the next period and the borrowing costs charged by the bank holding
company. If the loan rate is perfectly flexible (κb = 0), the maximization
problem simplifies to rbt = rwb
t µb/(µb − 1) and rbt = rwb
t . Hence each branch
simply sets the loan rate as a mark-up over its marginal costs. Note that
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 110PDF page: 110PDF page: 110PDF page: 110
100 Chapter 3
default losses do not affect the optimization problem of the retail branch, but
they do affect the rate set by the bank holding company, rwbt , bank holding
companies charge their retail branches. Hence, an increase in default losses
does affect rbt via an increase in rwb
t .
The solution to the funding branch optimization problem is:
rdt = ζd
1 rdt−1 + ζd
2Etrdt+1+ ζd
3 rt, (3.A.57)
where ζd1 ≡ κd/(µd + 1 + (1 + βH)κd), ζd
2 ≡ βHκd/(µd + 1 + (1 + βH)κd),
and ζd3 ≡ µd + 1/(µd + 1 + (1 + βH)κd). Similar to the lending branch, the
deposit rate depends on the deposit rate in the previous period, the deposit
rate in the coming period, and the lending rate offered by the bank holding
company. If the deposit rate is perfectly flexible (κd = 0) the maximization
problem simplifies to rdt = rtµd/(µd − 1) and rd
t = rt.
The central bank stabilizes the economy via a simple Taylor rule:
rt =δr rt−1 + (1 + r)(1− δr)δπ(πt + Etπt+1)+
(1 + r)(1− δy)δr(yt − yt−1) + ηmt . (3.A.58)
All that remains is the evolution of the state variables and the goods market
equilibrium that closes the model. The capital accumulation identities are:
kt =δk it + (1− δk)kt−1, (3.A.59)
kbt =kb
t−1 −δbb∗
kb∗ bt−1 +jb∗
kb∗ jbt +
t∗
kb∗ tt, (3.A.60)
where in steady state δbb/kb∗ = jb/kb∗ = δb/ν and we set t∗/kb∗ = δb/ν, i.e,
the bank recapitalization is added to bank profits.
As the changes in deposits and bonds are important for leverage in the
banking sector, these variables are modeled explicitly via the budget iden-
tities of the households and the entrepreneurs:
dt =(1 + rd)dt−1 + rdt−1 − πt +
w∗lh∗
d∗(wt +
ˆlHt − tt) +
rp∗
c∗rpt−
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 111PDF page: 111PDF page: 111PDF page: 111
Credit defaults and bank capital 101
cH∗
d∗cH
t , (3.A.61)
and
bt =−rkk∗
b∗([1− Γ(ω)]
(rk
t+1 + kt
)− Γ(ω)ω ˆωt+1
)+(
1 + rb[1− Γ(ω)])
bt−1 + [1− Γ(ω)]rb(
rbt−1 − πt
)−
Γ′(ω)rbω ˆωt+1 +
i∗
b∗it +
wlH∗
b∗(
wt + LHt + tt
)+
cE∗
b∗cE
t . (3.A.62)
As t∗ is in steady state equal to zero, we simply subtract the lump-sum tax
from labor income. Bank profits and expected credit default losses are rep-
resented by:
jbt =
rb
δb (bt + rbt )−
rd (1− ν)
δb (dt + rdt )−
rd (1− ν)−(δb − rb)
δb (vt + Etvt+1 − Et−1vt), (3.A.63)
and
vt =ξv∗(rbt + bt) + (1− ξv∗)(rk
t+1 + qt + kt)−
ξv∗(
f (ω)ω
2+ F(ω)
)ˆωt + ηv
t , (3.A.64)
where
ξv∗ ≡ F(ω∗)b∗rb∗
F(ω∗)b∗rb∗ − (1− µ)∫ ω∗
0 ω∗ f (ω∗)dω∗rk∗k∗
=F(ω∗)[Γ(ω∗)− µG(ω∗)]
F(ω∗)[Γ(ω∗)− µG(ω∗)]− (1− µ)G(ω∗). (3.A.65)
We used the steady state conditions for bank profits jb∗ = rbb∗ − rdd∗ − v∗,
bank capital accumulations δbb∗ = jb∗, bank balance sheet b∗ = d∗+ kb∗ and
kb∗/b∗ = ν. These conditions give 1− ν = d∗/b∗ and v∗/b∗ = rd (1− ν)−
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 112PDF page: 112PDF page: 112PDF page: 112
102 Chapter 3
(δb − rb). The model is closed by the goods market equilibrium condition:
yt =
(1− δk k∗
y∗
)ct + δk k∗
y∗it. (3.A.66)
Linearization of the countercyclical capital buffers yields:
myt = $y
(bt − yt
), (3.A.67)
mkb
t = $kb
(kb
t
), (3.A.68)
and once activated the lending rate set by the bank holding company is
determined by the whole sale bank as:
rwbt − rt =−
κwν3
r(kb
t − bt − myt ), (3.A.69)
or
rwbt − rt =−
κwν3
r(kb
t − bt − mkb
t ), (3.A.70)
depending on which activation rule the central bank applies.
3.B Variables
Consumption: Final consumption aggregates - Current prices seasonally
adjusted and adjusted data by working day in millions of euros. Source:
Eurostat.
Output: GDP and main components - Current prices seasonally adjusted
and adjusted data by working day in millions of euros. Source: Eurostat.
Investment: Gross fixed capital formation - Current prices seasonally adjus-
ted and adjusted data by working day in millions of euros. Source: Eurostat.
Wages: Gross wages and salaries seasonally adjusted and adjusted data by
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 113PDF page: 113PDF page: 113PDF page: 113
Credit defaults and bank capital 103
working day in millions of euros. Source: Eurostat.
Price inflation: Harmonized Index of Consumer Prices (2005=100) monthly
data. Source: Eurostat.
Nominal policy rate: Money market interest rates - monthly data (day-to-
day) Eurostat.
Outstanding loans to firms: Euro area (changing composition), Outstand-
ing amounts at the end of the period (stocks), MFIs excluding ESCB re-
porting sector - Loans, Total maturity, All currencies combined - Euro area
(changing composition) counterpart, Non-Financial corporations (S.11) sec-
tor, denominated in Euro, data neither seasonally nor working day adjusted.
Source: European Central Bank Statistical Data Warehouse.
Outstanding deposits to households: Euro area (changing composition),
Outstanding amounts at the end of the period (stocks), MFIs excluding ESCB
reporting sector - Overnight deposits, Total maturity, All currencies com-
bined - Euro area (changing composition) counterpart, Households and non-
profit institutions serving households (S.14 & S.15) sector, denominated in
Euro, data neither seasonally nor working day adjusted. Source: European
Central Bank Statistical Data Warehouse.
Nominal interest rate on loans: Euro area (changing composition), Annu-
alised agreed rate (AAR) / Narrowly defined effective rate (NDER), Credit
and other institutions (MFI except MMFs and central banks) reporting sector
- Loans other than revolving loans and overdrafts, convenience and exten-
ded credit card debt [A20-A2Z], Over 1 and up to 5 years initial rate fixa-
tion, Over EUR 1 million amount, New business coverage, Non-Financial
corporations (S.11) sector, Euro. Source: European Central Bank Statistical
Data Warehouse.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 114PDF page: 114PDF page: 114PDF page: 114
104 Chapter 3
Nominal interest on deposits: Euro area (changing composition), Annual-
ised agreed rate (AAR) / Narrowly defined effective rate (NDER), Credit
and other institutions (MFI except MMFs and central banks) reporting sec-
tor - Deposits with agreed maturity, Up to two years original maturity, New
business coverage, Households and non-profit institutions serving house-
holds (S.14 and S.15) sector, Euro. Source: European Central Bank Statistical
Data Warehouse.
Credit risk: the credit risk indicator is constructed by Gilchrist and Mojon
(2017).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 115PDF page: 115PDF page: 115PDF page: 115
Chapter 4
Mortgage loans and shadow
banks
4.1 Introduction
Since the 1980s the banking sector has been characterized by three major
trends. Perhaps the best documented trend is the secular decline in (real)
interest rates, see Figure 4.1. In March 2005 former Fed chairman Bernanke
emphasized the existence of a global savings glut, an increase in the global
supply of savings, as the main source of this decline in interest rates.1 An
efficiently functioning banking sector allocates these savings to their most
productive use. However, starting in the 1980s and up to the recent global
financial crisis a second trend is observed. Banks increased the amount of
loans secured by real estate, henceforth mortgages, much faster than the
amount of commercial and industrial loans, henceforth corporate loans, see
Figure 4.2. This reallocation of bank lending was accompanied by a third
trend: a shift from regulated banking towards unregulated banking (hence-
forth: shadow banking), see Figure 4.3.
There is a clear correlation between the growth of shadow bank liabil-
1 Bernanke (2005) argues that developing countries increased their savings rate by decreas-ing their consumption. These savings were subsequently used to buy assets from developedcountries which put downward pressure on their interest rates. See also Caballero et al.(2008) for a more detailed analysis of the decline in real interest rates.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 116PDF page: 116PDF page: 116PDF page: 116
106 Chapter 4
ities and the growth rate of mortgage loans, see Figure 4.4. However, no
encompassing framework exists that describes a causal relationship in a
general equilibrium context as these trends have, typically, not been con-
sidered together. Nevertheless, recent concerns regarding financial stability
imply that such a framework is needed. These concerns arise because, on
the one hand, a reallocation of lending towards housing investment rather
than physical capital formation might harm the economy’s production capa-
city, making it eventually harder to repay mortgage debt. A vastly growing
shadow banking sector, on the other hand, might increase the likelihood of
fire-sales and bank runs.2 This chapter builds a tractable model that shows
how an exogenous inflow of deposits on bank balance sheets depresses real
interest rates economy-wide and increases the share of mortgages on the ag-
gregate bank balance sheet. This in turn fosters growth of, in particular, the
shadow banking sector.3 In doing so, we show that growth of the shadow
banking sector reduces financial stability because shadow banks create more
uninsured deposits than socially optimal.
The intuition behind the relative reallocation of bank lending from cor-
porate loans to mortgage loans is as follows. The model distinguishes two
assets that can serve as collateral for bank loans: houses for mortgages and
physical capital for corporate loans. By assumption the assets differ with
respect to their supply elasticity; on average the supply of physical capital
is more elastic than the supply of houses. An exogenous inflow of depos-
its on bank balance sheets depresses interest rates and induces impatient
households to demand more credit for both houses and physical capital. In-
elastic housing supply relative to the supply of physical capital causes house
2 For example Bernanke (2005) argues that in the long run housing adds less to productivitygrowth than productive capital. Mian and Sufi (2014) and Mian et al. (2016) show that anincrease in the household debt-to-GDP ratio predicts lower GDP growth and higher unem-ployment in the succeeding periods. Gennaioli et al. (2013) show how an exogenous increasein savings drives securitization, leverage and financial instability in the shadow banking sec-tor. Moreira and Savov (2014) examine the interaction between shadow banks and the realeconomy and show how shadow banks can create liquidity via securitization but also addi-tional instability.
3 Thereby the model also provides a theoretical explanation for the empirical findings ofJorda et al. (2015) who show that the rise in mortgage debt was closely associated with loosemonetary policy.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 117PDF page: 117PDF page: 117PDF page: 117
Mortgage loans and shadow banks 107
Figure 4.1. Up and downward sloping trend in effective federal funds rate
0
2
4
6
8
10
12
14
16
18
20
1957 1962 1967 1972 1977 1982 1987 1992 1997 2002 2007 2012 2017
Per
cen
tag
e p
oin
ts
Time (years)
Recession Effective Federal Funds Rate
The effective federal funds rate is the interest rate at which depository institutions tradefederal funds (balances held at Federal Reserve Banks) with each other overnight. Source:Federal Reserve Bank of St. Louis database (FRED Economic Data).
Figure 4.2. Real estate and corporate loans as share of U.S. financial sector
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
1934 1942 1950 1958 1966 1974 1982 1990 1998 2006 2014
Rel
ati
ve
to t
ota
l fi
na
nci
al
ass
ets
Time (year)
Recessions Loans secured by real estate Commerical and industrial loans
Notes: The blue line denotes loans secured by real estate (predominately mortgage loans).The red line denotes commercial and industrial loans (corporate loans). Both as a share of thetotal amount of financial assets of the domestic financial sector in the U.S. Source: Historicalstatistics on banking (Federal Deposit Insurance Corporation).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 118PDF page: 118PDF page: 118PDF page: 118
108 Chapter 4
Figure 4.3. Assets regulated and shadow banks as % of U.S. financial sector
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
1957 1962 1967 1972 1977 1982 1987 1992 1997 2002 2007 2012 2017
Rel
ati
ve
to T
ota
l F
ina
nci
al A
sset
s
Time (years)
Recession Fraction of Shadow Banks Fraction of Regulated Banks
Notes: The blue line shows the total liabilities of private depository institutions (regulatedregulated banks). The red line shows other financial intermediaries except insurance com-panies and pension funds (unregulated shadow banks). Both as a share of the total amountof financial assets in the U.S. Source: Flow of funds accounts of the United States (FRB).
Figure 4.4. Growth of mortgage loans and shadow bank liabilities U.S.
-15%
-10%
-5%
0%
5%
10%
15%
20%
25%
30%
35%
1957 1962 1967 1972 1977 1982 1987 1992 1997 2002 2007 2012 2017
Yea
r-o
n-Y
ear
cha
ng
es (
%)
Time (years)
Recession Growth Shadow Bank Assets Growth Mortgage Loans
Notes: the red lines shows the growth rate of other financial intermediaries (unregulatedshadow banks). The blue line shows the growth rate mortgage loans. Correlation 0.69.Source: Author’s own calculations using data from the Flow of funds accounts of the UnitedStates and Federal Reserve Bank of St. Louis database.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 119PDF page: 119PDF page: 119PDF page: 119
Mortgage loans and shadow banks 109
prices to rise relative to the price of physical capital. The value of collateral
for mortgage loans increases relative to the value of collateral for corporate
loans. Consequently, banks increase mortgage lending relative to corporate
lending. Hence, collateral with a low supply elasticity shows more price and
credit supply fluctuations than collateral with a high supply elasticity.
After having established a relationship between an inflow of deposits
and a relative increase in mortgage lending, we show how this relation-
ship can foster growth of, in particular, the shadow banking sector. In the
model, both regulated and shadow banks can create safe money-like claims
(deposits) which allows them to extract a rent from households, see, e.g.,
Gorton and Pennacchi (1990), Stein (2012) and Krishnamurthy and Vissing-
Jorgensen (2015). To keep deposits perfectly safe and liquid in any state of
the world, regulated banks are compelled to buy deposit insurance. As in
Stein (2012) and Hanson et al. (2015), shadow banks are not compelled to
participate in the deposit insurance scheme. To create ex-ante equally safe
and liquid deposits, shadow banks offer depositors an early liquidation op-
tion. Consequently, shadow banks have lower funding costs (regulatory ar-
bitrage), but are exposed to liquidity risk and prefer to hold highly liquid
assets.
In the model, all banks trade in the interbank market (e.g., to construct
diversified portfolios or because they face idiosyncratic liquidity shocks).
Consequently, when the aggregate banking sector grows, interbank trading
increases, the interbank market appears deeper and market liquidity—the
ease to sell an asset close to its fundamental value—increases, see also Han-
son et al. (2015). As the inflow of deposits fosters growth of mortgage loans,
the interbank market for mortgage loans becomes deeper and the expec-
ted liquidation value of selling mortgage loans increases. Since the expected
liquidation value of shadow bank assets increases, their funding liquidity
increases. Shadow banks face less liquidity risk and can exploit their regu-
latory arbitrage with respect to the supply of mortgage loans. As a result,
the shadow banking sector grows relative to the regulated banking sector
because the mortgage loan market grows relative to the corporate loan mar-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 120PDF page: 120PDF page: 120PDF page: 120
110 Chapter 4
ket.
The growing shadow banking sector leaves the economy excessively
vulnerable to financial crises because the shadow banking sector creates too
many uninsured deposits compared to the social optimum. As in Brunner-
meier and Pedersen (2009) and Stein (2012), when intermediaries do not
internalize the costs of fire-sales, unregulated private money creation is typ-
ically sub-optimal. In this chapter, we utilize the general equilibrium struc-
ture of the model to show why banks do not internalize this price effect. A
growing shadow banking sector increases the risk that a significant share of
the banking sector must liquidate its assets which reduces liquidity in the
interbank market. Consequently, shadow banks that create more uninsured
deposits increase the banking system’s reliance on liquidity support by the
central bank. However, the expected costs of liquidity support by the central
bank are not affected by the size of the shadow banking sector. As a result,
the increase in liquidity risk is not fully reflected in the expected liquidation
price of shadow banks’ assets and they issue too many uninsured deposits.4
The results presented in this chapter relate to a growing literature em-
phasizing the risks for financial and economic stability of household debt.
First, an inflow of deposits might harm production when mortgage loan
supply crowds out corporate loan supply as investment in physical capital
is more productive than investment in housing, see Benigno and Fornaro
(2014), Bernanke (2005) and Borio et al. (2016). In the model, the total stock
of houses is fixed.5 Hence, the increase in mortgage lending is collateral-
ized by the increase in the price of the underlying asset rather than by an
increase in the total amount of underlying assets. That is, the economy is
increasingly Ponzi-financed (Minsky, 1986). The increase in mortgage loans
supported solely by an increase in house prices does not improve the eco-
4 Mink (2016) and DeAngelo and Stulz (2015) show that banks having access to a very liquidinterbank market in which systemic risk is insured by a lender of last resort leads to excessiveliquidity creation and high bank leverage.
5 Green et al. (2005) argue that housing supply responds relatively inelastic to increasesin housing demand in the short-run. Saiz (2010) shows that most areas in which housingsupply is regarded as inelastic are constrained by the amount of available land related to thegeography of the area.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 121PDF page: 121PDF page: 121PDF page: 121
Mortgage loans and shadow banks 111
nomy’s production capacity and leaves the economy vulnerable to future
house price declines.6 Empirical evidence reported by Mian and Sufi (2014)
and Mian et al. (2016) shows indeed that household debt, predominantly
mortgage loans, is an important determinant of business cycle fluctuations.
Prudential authorities in some countries introduced restrictions on ad-
missible loan-to-value (LTV) ratios to create a precautionary buffer against
house price fluctuations. Here we show two additional potential benefits
from these LTV limits. First, they mitigate house price and thereby mortgage
supply fluctuations when shocks hit the economy. Hence, LTV limits do not
only provide a larger buffer to protect borrowers against house price fluc-
tuations but, also simultaneously attenuate fluctuations in house prices and
mortgage supply. These results are in line with the findings of Wong et al.
(2011) which emphasize the importance of LTV caps in reducing systemic
risk originating from the boom-and-bust cycle of housing markets. Second,
LTV limits reallocate bank lending from mortgage loans to corporate loans,
not only in steady state but also when interest rates fall and credit supply in-
creases. As a result, a larger share of the increase in credit supply is allocated
to firms investing in physical capital which benefits the economy’s produc-
tion capacity. These results emphasize the importance of macroprudential
tools like LTV limits in safeguarding financial and economic stability.
This chapter is also related to a growing literature on financial stability.
Absent a deposit guarantee system, a large shadow banking sector under-
mines financial and economic stability. As described by Diamond and Dyb-
vig (1983), the liquid character of bank liabilities and the illiquid character
of bank assets make these banks intrinsically prone to runs. Brunnermeier
(2009), Brunnermeier and Pedersen (2009) and Brunnermeier and Sannikov
(2014) describe how banks that fund themselves with liquid short-term de-
posits to finance illiquid long-term investments are more vulnerable to li-
quidity risk. Pozsar et al. (2010) and Adrian and Ashcraft (2012) argue that
deposit insurance and a liquidity backstop provided by the central bank are
6 Glaeser et al. (2008) show indeed that places with more elastic housing supply have fewerand shorter house price bubbles.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 122PDF page: 122PDF page: 122PDF page: 122
112 Chapter 4
key to keep consumers calm and their claims liquid. However, by providing
liquidity insurance to banks that is not actuarially fair priced, banks have a
competitive advantage in creating risk free and liquid claims. It is, precisely
this free liquidity insurance that induces shadow banks to issue too many
deposits rather than raising equity, leaving the financial sector vulnerable to
liquidity crises.7
In theory this externality and thereby the consequences for financial sta-
bility could be regulated by means of a Pigouvian tax. In practice, how-
ever, it is hard to determine the optimal level of this tax. Stein (2012) sug-
gests therefore that the regulator should supply tradable permits that allow
banks to create deposits. The market price of these tradable permits helps
to identify the externality. Nevertheless, the optimal amount of permits re-
mains a guess and, if anything, the system increases regulatory arbitrage
between regulated and unregulated banks. Here, we propose a more direct
approach that does not rely on unobservables and affects shadow banks and
regulated banks equally as it works through market prices. The central bank
should provide households access to interest-paying central bank deposits.
In doing so, the opportunity costs for households of holding bank deposits,
as opposed to interest bearing central bank deposits, increases. This means
that banks have to offer a higher interest rate on deposits to persuade house-
holds not to convert their deposits into central bank deposits. If the interest
rate on central bank deposits is sufficiently high, the central bank eliminates
the incentive for both regulated and shadow banks to finance themselves
with deposits rather than equity.
The rest of the chapter is organized as follows. Section 4.2 describes the
correlation presented in Figure 4.4 in more detail. Section 4.3 presents the
model. Section 4.4 presents the simulation results. Section 4.5 discusses the
policy implications and Section 4.6 concludes.
7 In this chapter we use a macro-perspective to motivate the externality. Stein (2012), Green-wood et al. (2015) and Greenwood et al. (2016) motivate a similar externality from the per-spective of an individual bank and also argue that in its presence banks create too manyuninsured deposits.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 123PDF page: 123PDF page: 123PDF page: 123
Mortgage loans and shadow banks 113
4.2 Stylized facts extended
In this section we examine more carefully which factors drive the correlation
presented in Figure 4.4. Specifically, we control for real business cycle activ-
ity to ensure that the correlation in Figure 4.4 is not driven by an omitted
variable or underlying trend. Also, Gertler et al. (2016) suggest that finan-
cial innovation (mostly securitization of mortgages) was the main driver of
shadow bank growth which led to more leverage capacity at shadow banks,
a decrease in lending rates and more mortgage lending. First, we control
for financial innovation by including the growth rate of the asset backed
securities and mortgage backed securities markets. Second, we estimate a
vector autoregression in an attempt to control for reverse causality. Finally,
the correlation in Figure 4.4 is biased when the growth rate of mortgage
loans is by definition equal to the growth rate of the shadow bank balance
sheet. Although we will argue that this form of simultaneity is unlikely, we
replace the growth rate of mortgage loans by the growth rate of mortgage
loans supplied by private depository institutions only. This growth rate is
less likely to by determined simultaneously with the growth rate of shadow
bank assets.
4.2.1 Linear regression model
To examine more carefully whether the correlation is caused by any of these
factors, we estimate the following linear regression model for the U.S. bank-
ing sector:
∆ ln (Ybt ) = α0 + α1∆ ln (Mtotal
t ) + α2∆ ln (Ctotalt ) + βBreal
t + γZtotalt + εt
(4.1)
where Ybt denotes either the total amount of regulated (rb) or shadow bank
(sb) financial assets, b ∈ (sb, rb), Mtotalt denotes the total amount of mort-
gage loans and Ctotalt denotes the total amount of corporate loans in the U.S.,
Brealt controls for real business cycle activity (real GDP growth, the Federal
Funds rate and CPI inflation) and Ztotalt controls for financial innovation and
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 124PDF page: 124PDF page: 124PDF page: 124
114 Chapter 4
includes both the growth rate of the total asset backed securities (ABS) and
mortgage backed securities (MBS) markets in the U.S., see Tables 4.A.1 and
4.A.2 in Appendix 4.A for details.8
The coefficients in Equation (4.1) are subject to a simultaneity bias when
the growth rate of mortgage loans and the growth rate of corporate loans
determine, to a large extent, the growth rate of the bank balance sheet Ybt .
However, as shadow banks and regulated banks only constitute a relatively
small share of the total U.S. financial sector, this bias is arguably small. These
shares for shadow banks and regulated banks range from 5 and 42 percent in
1954 to 22 and 18 percent in 2017, respectively (see also Figure 4.3). In addi-
tion, mortgage loans account for (on average) 22 percent of the total amount
of financial assets in the U.S while corporate loans account for only 6 percent
of the total amount of financial assets. About half of these mortgage loans
(48 percent) are supplied by private depository institutions, i.e., regulated
banks. To estimate whether a simultaneity bias is driving our results, we
also estimate Equation (4.1) for shadow banks and replace the growth rate
of the total amount of mortgage loans, by the growth rate of mortgage loans
supplied by private depository institutions only.
We estimate Equation (4.1) with Ordinary Least Squares. The results are
presented in Table 4.1. Columns (1) and (2) present the estimation results for
the growth rate of regulated bank assets. All regressors are significant and
have the expected sign. The coefficients on mortgage loan growth (α1) and
corporate loan growth (α2) are not significantly different from each other.
These results suggest that both types of assets are equally important for the
growth rate of regulated banks. The growth rate of the ABS and MBS mar-
kets (γ = 0.089) correlates significantly with the growth rate of regulated
banks. The coefficient is, however, relatively small and suggests that growth
of the ABS and MBS markets is not an important factor for the growth rate
of the regulated banking sector.
The results for shadow banks are presented in columns (3) and (4) of
Table 4.1. Shadow bank growth correlates strongly with the growth rate
8 All variables are stationary at the 5% significance level.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 125PDF page: 125PDF page: 125PDF page: 125
Mortgage loans and shadow banks 115
of mortgage loans (α1 = 0.646). Also the growth rate of corporate loans
is significantly correlated with the growth rate of shadow banks. However,
it appears that for shadow banks the growth rate of corporate loans is less
important. The size of the coefficient on corporate loan growth α2 = 0.176,
which is about one-fourth of the size of the coefficient on mortgage loan
growth α1. These results suggest that the growth rate of the mortgage loan
market is much more important for shadow bank growth than the growth
rate of the corporate loan market.
The growth rate of the market for ABS and MBS also significantly cor-
relates with the growth rate of the shadow banking sector. However, the
coefficient γ = 0.068 is again small. These results suggest that shadow bank
growth is not strongly correlated with the expansion of the ABS and MBS
market.
Finally, columns (5) and (6) present the results for shadow banks were
we have replaced the growth rate of mortgage loans by the growth rate of
mortgage loans supplied by private depository institutions only. Although
this variable does not capture the full dynamics of the mortgage loan mar-
ket, the coefficients are no longer subject to a simultaneity bias. The coef-
ficient on mortgage loan growth drops from 0.646 tot 0.378, which is still
significantly larger than the growth rate on corporate loan growth which is
0.175. These results suggest that an increase in the amount of mortgage loans
supplied by private depository institutions also correlates significantly with
the growth rate of shadow bank assets.
4.2.2 Multivariate regression model
For shadow banks, the coefficients in Table 4.1 on the real economic con-
trol variables (β1, β2, β3) are either insignificant or have an unexpected sign.
These coefficients can be biased due to reverse causality when the growth
rate of the shadow banking sector also affects the growth rate of GDP, the
inflation rate and the Federal Funds rate. In an attempt to control for this
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 126PDF page: 126PDF page: 126PDF page: 126
116 Chapter 4
Table 4.1. Estimation results Equation (4.1)
Regulated banks Shadow banks Shadow banks
Variables Coefficients Prob. Coefficients Prob. Coefficients Prob.(1) (2) (3) (4) (5) (6)
Growth of mortgages loans (α1) 0.279 0.000 0.646 0.000(0.044) (0.073)
Growth of mortgages loans 0.378 0.000PDI only (α1) (0.065)
Growth corporate loans (α2) 0.215 0.000 0.176 0.000 0.175 0.001(0.026) (0.044) (0.053)
GDP Growth (β1) 0.153 0.000 -0.093 0.180 -0.079 0.302(0.042) (0.070) (0.076)
Federal Funds Rate (β2) -0.328 0.000 0.852 0.000 1.087 0.000(0.072) (0.119) (0.132)
Inflation (CPI) (β3) 0.358 0.000 -0.324 0.018 -0.352 0.018(0.082) (0.135) (0.147)
Growth ABS & MBS assets (γ) 0.089 0.000 0.058 0.035 0.082 0.005(0.016) (0.027) (0.029)
Constant (α0) 2.086 0.000 1.016 0.068 1.934 0.001(0.334) (0.554) (0.576)
Notes: Estimation period 1955Q3 − 2017Q2. Standard errors in parentheses Adjusted R-squared: 0.604 for the growth rate of regulated banks and 0.705 and 0.705 for the growthrate of shadow banks. Abbreviation: Private Depository Institutions (PDI). Source: Author’sown calculations using data from the Flow of funds accounts of the United States and Fed-eral Reserve Bank of St. Louis database.
reverse causality we estimate the following vector autoregression:9
Xt =α0 + Φ(L)Xt−1 + εt, (4.2)
where Φ(L) ≡ Φ0 +Φ1L1 + ...+ΦpLp is a lag polynomial and Xt is a stacked
vector containing the same observed variables as in (4.1). We identify the
VAR by using a Cholesky decomposition. The ordering is as follows: Xt =
[Ybt , Mtotal
t , Ctotalt , Breal
t , Ztotalt ] and we include 4 lags. Results are robust to dif-
ferent lag lengths. Also the ordering of the VAR does not affect the res-
ults. Figure 4.7 in Appendix 4.A shows the results for a different order-
ing: Xt = [Mtotalt , Ctotal
t , Ybt , Breal
t , Ztotalt ], i.e., we allow the growth rate of the
shadow banking sector to affect the growth rate of the mortgage loan mar-
ket and the corporate loan market contemporaneously. Results are very sim-
9 The series do show some signs of co-integration. For this reason we also estimated a VectorError Correction model. The results, which are not presented here, are similar to the impulseresponse functions of the Structural Vector Autoregression.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 127PDF page: 127PDF page: 127PDF page: 127
Mortgage loans and shadow banks 117
ilar. The results are also robust to placing the real economic variables Brealt
and/or the growth rate of the ABS and MBS markets Ztotalt first in order.
Figure 4.5 shows the impulse response functions for regulated banks. For
brevity we omitted the impulse response functions for the other variables in-
cluded in Equation (4.2). The graphs shown in the first column show the re-
sponse of mortgage loan growth and corporate loan growth to an unexpec-
ted increase in the growth rate of the regulated banking sector. The growth
rate of both assets increases when the regulated banking sectors grows. The
growth rate of corporate loans responds stronger than the growth rate of
mortgage loans to an increase in the growth rate of regulated banks. The res-
ults suggest that a one percentage point increase in regulated bank growth
is related to a 1 percentage point increase in corporate loan growth, and only
to a 0.5 percentage point increase in mortgage loan growth.
The first graphs in the second and third column show the response of the
growth rate of the regulated banking sector to a shock in the growth rate of
mortgage loans and corporate loans, respectively. Both impulse responses
are insignificant suggesting that the growth rate of regulated banks is not
affected by unexpected increases in the growth rates of these assets. That is,
we cannot find evidence that an increase in the markets for corporate loans
or mortgage loans affects the growth rate of the regulated banking sector.
Figure 4.6 shows similar impulse response functions as in Figure 4.5, but
replaces the growth rate of the regulated banking sector with the growth
rate of the shadow banking sector. It turns out that the results for shadow
banks are different than those for regulated banks. Shadow bank growth
appears to respond positively to a shock in the mortgage growth rate (first
graph in the second column). In contrast, shadow bank growth is unaffected
by an increase in the growth rate of corporate loans (first graph in the third
column). Furthermore, an increase in the growth rate of shadow banks has
no effect on the growth rate of either mortgage loans or corporate loans.
Figure 4.7 shows similar impulse response functions as in Figure 4.6, but
replaces the growth rate of mortgage loans with the growth rate of mortgage
loans supplied by private depository institutions. Results are very similar to
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 128PDF page: 128PDF page: 128PDF page: 128
118 Chapter 4
Figure 4.5. VAR estimation results of Equation (4.2) for regulated banks
Corporate loan to regulated bank Corporate loan to mortgage loan Corporate loan to corporate loan
Note: time (year) on the horizontal axis, percentage point devations on the vertical axis.
Regulated bank to Regulated bank Regulated bank to mortgage loan Regulated bank to corporate loans
Mortgage loan to regulated bank Mortgage loan to mortgage loan Mortgage loan to corporate loan
-1
0
1
2
1 2 3 4 5 6 7 8 9 10-1
0
1
2
1 2 3 4 5 6 7 8 9 10-1
0
1
2
1 2 3 4 5 6 7 8 9 10
-1.0
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
-1.0
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
-1.0
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10
-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10
Notes: estimation period 1955Q3–2017Q2, time (quarters) horizontal axis, percentage pointdeviation on the vertical axis. The titles explain the response of one of the variables includedto a shock of any of the variables included. Dotted red lines denote confidence intervals at95% significance level.
the results presented in Figure 4.6: a shock to the growth rate of mortgage
loans affects the growth rate of the shadow banking sector, whereas an in-
crease in the shadow bank growth rate does not affect the growth rate of
mortgage loans. Moreover, replacing the the growth rate of mortgage loans
with the growth rate of mortgage loans supplied by private depository in-
stitutions in reduced the size of the coefficient significantly in Table 4.1. In
contrast, the results in 4.7 are not affected by this change and appear robust.
Evidently these results are not conclusive. They do suggest, however,
that mortgage growth appears to affect shadow bank growth, while we can-
not find strong evidence that shadow bank growth is driving the growth
rate of the mortgage loan market. In the next section we present a model
that is able to explain this apparent causality in more detail.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 129PDF page: 129PDF page: 129PDF page: 129
Mortgage loans and shadow banks 119
Figure 4.6. VAR estimation results of Equation (4.2) for shadow banks
Corporate loan to shadow bank Corporate loan to mortgage loan Corporate loan to corporate loan
Note: time (year) on the horizontal axis, percentage point devations on the vertical axis.
Shadow bank to shadow bank Shadow bank to mortgage loan Shadow bank to corporate loans
Mortgage loan to shadow bank Mortgage loan to mortgage loan Mortgage loan to corporate loan
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10
-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10
-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10
Response of GROWTH_SB to GROWTH_SB
Response to Cholesky One S.D. Innovations ± 2 S.E.
-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10
Response of GROWTH_SB to GROWTH_M
Response to Cholesky One S.D. Innovations ± 2 S.E.
-2
-1
0
1
2
3
1 2 3 4 5 6 7 8 9 10
Notes: estimation period 1955Q3–2017Q2, time (quarters) horizontal axis, percentage pointdeviation on the vertical axis. The titles explain the response of one of the variables includedto a shock of any of the variables included. Dotted red lines denote confidence intervals at95% significance level.
Figure 4.7. VAR estimation results of Equation (4.2) for shadow banks andPDI mortgage loan growth
Corporate loan to shadow bank Corporate loan to mortgage loan Corporate loan to corporate loan
Note: time (year) on the horizontal axis, percentage point devations on the vertical axis.
Shadow bank to shadow bank Shadow bank to mortgage loan Shadow bank to corporate loans
Mortgage loan to shadow bank Mortgage loan to mortgage loan Mortgage loan to corporate loan
-1
0
1
2
3
2 4 6 8 10
-1
0
1
2
3
2 4 6 8 10
-1
0
1
2
3
2 4 6 8 10
-2
-1
0
1
2
2 4 6 8 10
-2
-1
0
1
2
2 4 6 8 10
-2
-1
0
1
2
2 4 6 8 10
-2
0
2
4
2 4 6 8 10
-2
0
2
4
2 4 6 8 10-2
0
2
4
1 2 3 4 5 6 7 8 9 10
Notes: estimation period 1955Q3–2017Q2, time (quarters) horizontal axis, percentage pointdeviation on the vertical axis. The titles explain the response of one of the variables includedto a shock of any of the variables included. Dotted red lines denote confidence intervals at95% significance level.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 130PDF page: 130PDF page: 130PDF page: 130
120 Chapter 4
4.3 The model
The model embeds elements of the financial structure developed in Stein
(2012), Gennaioli et al. (2013) and Hanson et al. (2015) in a general equi-
librium context. Specifically, regulated banks are regulated while shadow
banks are not. Since households value deposits for their safety and liquidity,
the deposit rate trades at a discount vis-a-vis the rate on bank equity. Con-
sequently, absent regulation, both shadow banks and regulated banks prefer
to finance their lending with deposits rather than equity as the latter is more
expensive. Both banking types can fund mortgages and corporate loans. In
this chapter, corporate loans are used by firms to buy physical capital and
mortgage loans are used by impatient households to fund a house.
Aggregate productivity risk is introduced to create scope for deposit in-
surance and liquidity risk. Aggregate risk cannot be diversified and poses a
potential threat to the risk-free claim of depositors. The central bank requires
regulated banks to insure all downside aggregate risk in the deposit guar-
antee scheme (DGS). Shadow banks are unregulated and offer households
an early liquidation option to create risk-free claims (Stein 2012 and Hanson
et al. 2015). For shadow banks the early liquidation option is less costly, but
offering this option creates additional liquidity risk. Unregulated shadow
banks cannot borrow directly from the central bank. Consequently, if a pess-
imistic signal about the future state of the world occurs, shadow banks must
sell their assets in the interbank market to regulated banks in order to remu-
nerate their depositors. The expected liquidation value of shadow bank as-
sets is key and determines the shadow banks’ comparative advantage vis-a-
vis regulated banks. If the expected liquidation value is low, shadow banks
have no comparative advantage and only regulated banks exist. If the expec-
ted liquidation value is high, shadow banks have a comparative advantage
and can exploit any regulatory arbitrage.
The model describes three sources of heterogeneity: patient (p) and im-
patient (i) consumers denoted by the superscript j ∈ p, i, regulated banks
(rb) and shadow banks (sb) denoted by the superscript b ∈ rb, sb, and
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 131PDF page: 131PDF page: 131PDF page: 131
Mortgage loans and shadow banks 121
mortgages ( f ) and corporate loans (e) denoted by the superscript ι ∈ f , e.Impatient consumers discount the future more heavily than patient con-
sumers. As a consequence, the latter will prefer to save while impatient
consumers will prefer to borrow. Both households may hold central bank
money (henceforth: cash) which does not pay interest or buy risk-free money-
like claims (henceforth: deposits) from risk neutral banks because both are
necessary to consume. Thus, indirectly, consumption is subject to a cash-in-
advance constraint. Patient households might also buy equity of both bank-
ing types in order to save part of their income.
4.3.1 Aggregate risk
The return on bank assets (corporate loans and mortgages) πs depends on
the state of the world s ∈ S. At time t all banks assume that state s ∈ S
materializes with probability $s > 0 where ∑s $s = 1. For reasons of tract-
ability, we assume that at each point in time banks expect only 3 different
states: S = g, b, d referring to a good, bad and disaster state, respectively.
Accordingly, πg(·) > πb(·) > πd(·) denote the return on bank assets in the
respective states. Between time t and t + 1 we assume that some informa-
tion about the future economic state can be observed S′= H, L, which can
be either optimistic (H) or pessimistic (L).The probability tree in Figure 4.8
formalizes the discussion where P(u) denotes the probability of an up node
and P(d) denotes the probability of a down node. If an optimistic signal is
observed, agents know with certainty that the disaster state will not occur.
If a pessimistic signal is observed, all three states can occur.10
The following notation is introduced for brevity purposes:
Et(πs) ≡ P(u)Et|S′=H(πs) + P(d)Et|S′=L(πs) = 1, (4.3)
Et|S′=H(πs) ≡ P(u|H)πg + P(d|H)πb, (4.4)
10 These assumptions can easily be generalized. In particular, S could include a continuousset of states s and a continuous set of signals might be observed that are consistent withthe results presented below. Crucial, however, is that a set of signals can be observed withprobabilities of a disaster states that are negligibly small (or absent) to keep depositors calmwhile pessimistic signals result in depositors withdrawing their shadow bank deposits.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 132PDF page: 132PDF page: 132PDF page: 132
122 Chapter 4
Figure 4.8. Probability tree for the occurrence of a signal and the states of theworld.
t
L
d : πdt + 1
P(d|L)
b : πb1− P(u|L)− P(d|L)
g : πg
P(u|L)
P(d)
H
P(d|H)
P(u|H)
P(u)
Notes: P(u) denotes the probability for an up node and P(d) denotes the probability of adown node. After the signal, the probabilities of an up node or down node are conditionalon an optimistic (H) or pessimistic (L) signal. πg(·) > πb(·) > πd(·) denote the productivityin the good, bad and disaster state, respectively.
Et|S′=L(πs) ≡ P(u|L)πg + (1− P(u|L)− P(d|L))πb + P(d|L)πd, (4.5)
where Et(πs) is the expected return at time t before the signal occurs which
we normalize to unity and Et|S′=H(πs) and Et|S′=L(πs) denote the expected
return associated with an optimistic or a pessimistic signal, respectively.
4.3.2 Real economy: households and firms
The economy consists of two types of infinitely lived households, the only
difference being that the impatient household has a lower discount factor
than the patient household βi < βp. Consequently, in equilibrium patient
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 133PDF page: 133PDF page: 133PDF page: 133
Mortgage loans and shadow banks 123
households save, while impatient households borrow. The economy occu-
pies a continuum of households with their mass normalized to unity. House-
holds derive utility from consumption Cjt , holding deposits or cash Mj
t, leis-
ure (1− Ljt) and owning a house H j
t .11 The corresponding household utility
function takes the following functional form:
U jt = Et
∞
∑t=0
(βj)t
[(Cj
t)η(H j
t)1−η]1−σc
1− σc +γm(Mj
t)1−σm
1− σm −γ
jl(Lj
t)1+σl
1 + σl
,
(4.6)
where deposits Mjt can be supplied by regulated banks Mj,rb
t and shadow
banks Mj,sbt or by the central bank Mj,cb
t in the form of cash, Et is the expect-
ation operator, 1− η denotes the weight of housing, σc, σm and σl denote,
respectively, the inverse of the elasticities with respect to consumption, de-
posits and work-effort, γm and γjl denote the weights of deposits and labor
relative to consumption in the utility function.
Residential houses play a key role in the model. In Equation (4.6) we as-
sume that households consume a consumption bundle that contains regular
consumption goods and their stock of houses. Consequently, if consumption
increases, housing demand increases because consumers equate the mar-
ginal rate of substitution to the price differential between house prices (qht )
and consumer prices (in this model normalized to unity). However, as res-
idential houses do not perish each period like consumption goods, owning
a house also yields a potential capital gain rht ≡ qh
t /qht−1 − 1. Accordingly,
both a decrease in the lending rate or an expected increase in house prices
increase housing demand.
If households receive a pessimistic signal, they liquidate their deposits
in the shadow banks and convert it in regulated bank deposits or cash. If the
11 Deposits in the utility function can be motivated by a cash-in-advance constraint as house-holds need deposits to pay for consumption. While deposits can be used directly for pay-ment, equity must first be converted into a liquid asset like deposits before a household canuse it for payment. This liquidity difference motivates why, apart from safety reasons, bothtypes of households value deposit holdings over equity when interest rates on both assetsare the same.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 134PDF page: 134PDF page: 134PDF page: 134
124 Chapter 4
signal is optimistic, households remain calm and hold on to their deposits in
the shadow banks. The option to liquidate at the intermediate stage makes
the shadow bank deposits ex-ante perfectly safe and liquid. Regulated bank
deposits are fully protected by the deposit guarantee system (DGS). Hence,
in the steady state households do not hold cash because the central bank
does not pay interest and cash is therefore inferior to deposits. Cash becomes
attractive, however, when deposit interest rates fall sufficiently (below zero)
or when the regulated banking sector is no longer able to create perfectly
safe and liquid claims.
Households maximize their utility subject to their budget constraints.
The patient household’s budget constraint is represented by:
Mpt + Qt + qh
t (Hpt − Hp
t−1) = (1 + imt−1)Mp
t−1 − imt−1Mp,cb
t−1+
(1 + iqt−1)Qt−1 + rh
t Hpt−1 − Cp
t + WtLpt + θ(Πb
t + Πpt ) (4.7)
where Qt = Qrbt + Qsb
t denotes the equity stakes of the patient households in
the regulated and shadow banks respectively, imt , either im,rb
t or im,sbt , denote
the rates on regulated and shadow bank deposits respectively, and iqt , either
iq,rbt or iq,sb
t , denotes the risky interest rates on regulated and shadow banking
equity respectively.12 The term Wt denotes the wage rate patient households
receive for supplying labor to the firms and Πpt and Πb
t denote firm profits
and bank profits redistributed to households where θ denotes the share of
patient households in the economy.13 Finally, households own a housing
stock H jt and realize a return on the house defined by rh
t . We assume that
a house is perfectly divisible. Each period, households can therefore buy or
sell part of their house in line with their demand for housing for a price qht .
The impatient households have the same utility function as the patient
households. They own, however, potentially different assets because they
are the savers and therefore also the investors in the economy. Particularly,
impatient households have two investment opportunities for which they
12 Note that imt−1 Mp,cb
t−1 in (4.7) accounts for the fact that cash does not pay interest.13 As the economy is fully flexible, firms profits are always zero, while bank profits flow tothe bank equity holders.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 135PDF page: 135PDF page: 135PDF page: 135
Mortgage loans and shadow banks 125
can borrow: they can invest in physical capital Kt or in housing Ht. Physical
capital has a cost ret and yields a return rk
t while housing costs r ft and yields
a return rht . The impatient household’s budget constraint is represented by:
B ft + Be
t −Mit =qh
t (Hit − Hi
t−1) + (1 + i ft−1)B f
t−1 + (1 + iet−1)Be
t−1−
(1 + imt−1)Mi
t−1 + imt−1Mi,cb
t−1 − rht Hi
t−1 −WtLit − rk
t Kt−1−
(1− θ)(Πbt + Πp
t ) + Cit + It, (4.8)
where B ft denotes mortgages for the nominal value of the house qh
t Hit and Be
t
denotes corporate loans for the nominal value of physical capital qkt Kt−1, i f
t
denotes the gross lending rate for mortgages and iet denotes the gross lend-
ing rate for physical capital. Impatient households rent out their physical
capital stock to firms for a rental rate rkt . The physical capital stock accumu-
lates according to:
Kt+1 = Kt(1− δ) + It
(1− φ
2
(It
It−1− 1)2 )
, (4.9)
where It denotes net investment in the physical capital stock and δ denotes
the deprecation of the physical capital stock. The second term in round
brackets in (4.9) denotes physical capital adjustment costs. The parameter
φ which denotes the degree of adjustment costs, determines to some degree
shifts in demand between physical capital and housing as it determines the
elasticity of physical capital supply.
The loan (Bιt) cannot be larger than the real value of its collateral. Without
this constraint, impatient consumers would like to borrow indefinitely to
finance housing, physical capital, but also consumption. As the main pur-
pose is to distinguish between investment in two types of assets, we restrict
borrowing for consumption. For this reason, the following inequality con-
straints are postulated for impatient households, Appendix 4.B presents a
proof that (4.10) and (4.11) hold with equality:
qht Hi
t ≥ ζ f B ft , (4.10)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 136PDF page: 136PDF page: 136PDF page: 136
126 Chapter 4
qkt Kt−1 ≥ ζeB
e
t , (4.11)
where ζ f is a cap on admissible loan-to-value ratios for mortgage loans and
ζe is a cap on admissible loan-to-value ratios for corporate loans. Inequalities
(4.10) and (4.11) state that the amount borrowed for housing B ft and physical
capital Be
t cannot be larger than the value of the underlying asset that serves
as collateral qht Hi
t and qkt Kt−1 corrected for the loan-to-value limit.
Household preferences determine both the return on banking equity and
the deposit rate. Intermediation adds value as households are willing to pay
a premium for a safe and liquid asset that pays interest. Alternatively, house-
holds can hold cash supplied by the central bank, which is safe but does not
pay interest. Hence, as long as banks create completely safe deposits and
offer a positive return on deposits, households are willing to hold deposits
rather than cash. Banks have access to deposit insurance and the interbank
market to insure liquidity risk while households have no access to these in-
stitutions and as such cannot construct a diversified portfolio to eliminate
credit and liquidity risk. For this reason, households do not want to invest
their endowment directly in corporate loans or mortgages.
The option for households to convert deposits into cash is a source of ag-
gregate liquidity risk in the banking sector. Households only convert their
deposits into cash when regulated banks are no longer able to guarantee the
safety and liquidity of their deposits. This might happen when the interbank
market stops functioning properly, for example, because shadow banks trig-
ger a loss spiral. Regulated banks use the interbank market to diversify their
portfolio and insure idiosyncratic liquidity risk. Without interbank market
regulated banks loose their ability to create safe and liquid deposits and
the DGS buffers are not large enough to cover the increase in liquidity risk.
Consequently, households are more likely to convert their deposits into cash
which does not pay interest, but is completely safe and liquid. It is precisely
this mechanism we will use later on to motivate the increase in aggregate
liquidity risk when the shadow banking sector grows.
Patient households maximize (4.6) subject to (4.7) with respect to con-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 137PDF page: 137PDF page: 137PDF page: 137
Mortgage loans and shadow banks 127
sumption, deposits, labor supply, housing accommodation and bank equity.
Impatient households maximize (4.6) subject to (4.8), (4.9), (4.10) and (4.11)
by choosing consumption, housing accommodation, deposits, labor sup-
ply, mortgages, corporate loans, physical capital and investment, see Ap-
pendix 4.B for details. Consequently, and in contrast to Stein (2012), Gen-
naioli et al. (2013) and Hanson et al. (2015), depository funding, equity fund-
ing and household borrowing demand are all endogenously determined in
the model. Additionally, we model an explicit outside option that enables
households during financial stress to withdraw liquidity from the system
by increasing their holdings of cash.
Firms rent their physical capital from the impatient households and hire
labor from both types of households to minimize their production costs sub-
ject to the aggregate production technology:
Yt = At(Kt−1)1−α(Lt)
α, (4.12)
where Yt denotes production, Lt = Lpt + Li
t is the sum of patient and im-
patient household labor, At denotes an aggregate productivity index which
follows a stochastic process At = exp(ηat ), where ηa
t = ρaηat−1 + εa
t and εat is
an i.i.d. productivity shock ∼ (0, σa).
4.3.3 Regulated and shadow banks
In the model two types of banks exist: regulated banks that are regulated
and shadow banks that are unregulated. Regulated regulated banks are ob-
liged to insure their deposits in the DGS and have access to the central bank
lending facility. Regulated banks pay an actuarially fair price for this deposit
insurance that pays-off in the disaster state (Hanson et al., 2015). House-
holds can convert their regulated bank deposits into cash because regulated
banks can borrow from the central bank. Shadow banks do not particip-
ate in the deposit insurance scheme and cannot borrow from the central
bank. Consequently, shadow bank deposits cannot be converted into cash,
but only into regulated bank deposits. Therefore, if between t and t + 1 a
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 138PDF page: 138PDF page: 138PDF page: 138
128 Chapter 4
pessimistic signal about the future state of the world occurs, depositors li-
quidate their shadow bank deposits which are no longer risk-free. To repay
their depositors, the shadow banks must liquidate their assets in the interb-
ank market. The expected liquidation value determines how much safe and
liquid deposits a shadow bank can issue ex-ante and thereby determines
indirectly the shadow banks’ funding cost.
In the main text, only fully diversified portfolios without idiosyncratic
risk are considered. Appendix 4.E shows that both regulated banks and
shadow banks will always completely diversify their portfolio if diversifica-
tion costs are reasonably low. For example, when the interbank market func-
tions properly transaction costs are low and banks can easily trade assets to
have diversified portfolios. We therefore assume that a properly functioning
interbank market is a necessary condition for the banking sector to construct
a fully diversified portfolio. Without a properly functioning interbank mar-
ket, bank deposits are no longer perfectly safe and liquid, and households
might convert deposits into cash.
Regulated and shadow banks, denoted by the superscript b ∈ (sb, rb), re-
spectively, have the same objective function except for the deposit insurance
premium. Each period the banks issue Mbt units of deposits and Qb
t units of
equity and promise to repay (1+ im,bt )Mb
t and (1+ iq,bt )Qb
t in the next period.
These funds are used to lend mortgages B f ,bt and corporate loans Be,b
t . The
objective function for both banks can be described by:
Πbt = ie
t Be,bt + i f
t B f ,bt − im,b
t (Mbt + Mb,x
t )− iq,bt Qb
t − Ξb − ξDt, (4.13)
where i ft and ie
t are the expected returns on investment in mortgages and
corporate loans respectively, Mbt is the sum of patient Mp,b
t and impatient
household deposits Mi,bt , and Mb,x
t denote foreign deposits for which de-
mand is specified below. All off-steady state excess returns Πrbt and Πsb
t ac-
cumulate to the equity holders. The term Ξb denotes fixed costs and ensures
that excess returns are zero in equilibrium. Both the deposit rate and equity
rate are in equilibrium pinned down by household preferences while loan
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 139PDF page: 139PDF page: 139PDF page: 139
Mortgage loans and shadow banks 129
supply is restricted by the collateral constraints. Accordingly, lending rates
might differ from the banks’ weighted average costs of funding which yields
non-zero bank profits without the fixed costs term. The last term is the de-
posit guarantee premium Dt. For regulated banks ξ = 1 and for shadow
banks ξ = 0. The ex-ante actuarially fairly priced deposit guarantee pay-
ment is expressed as:
Dt = χ[(1 + i f
t )B f ,rbt + (1 + ie
t)Be,rbt
], (4.14)
where χ = P(d)P(d|L)(πb − πd) denotes the DGS cost per unit of invest-
ment in the risky assets. The deposit insurance needs to cover only the dif-
ference between the bad and disaster state as regulated banks holds suffi-
cient equity to remain solvent in the good and bad state. Both types of banks
satisfy the same budget constraint:
Be,bt + B f ,b
t + ξ (Dt + Rt) ≤ Qbt + Mb
t + Mb,xt , (4.15)
where Rt denotes deposits, or reserves, at the central bank. Only regulated
banks have access to central bank deposits. By assumption central bank
deposits do not pay interest and are completely safe. Moreover, regulated
banks are not required to hold these central bank deposits which ensures
that in equilibrium they will not lend from the central bank to hold central
bank deposits.
The market also requires banks to hold sufficient equity. The equity buf-
fer constraint states that in the worst possible state banks should be able to
repay their risk-free deposits. Since both types of banks adopt a different
business model with respect to the creation of risk free claims, the regulated
bank equity buffer constraint is different from the shadow bank equity buf-
fer constraint. The regulated bank equity buffer constraint (or risk-weighted
assets constraint) is specified by:
Rrbt + πb
[(1 + i f
t )B f ,rbt + (1 + ie
t)Be,rbt
]≥ (1 + im,rb
t )(Mrbt + Mrb,x
t ).
(4.16)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 140PDF page: 140PDF page: 140PDF page: 140
130 Chapter 4
The buffer constraint states that regulated banks should have sufficient equity
to ensure that the return in the bad state of the world is sufficient to cover all
risk-free deposits. Deposit insurance guarantees the repayment of deposits
in the disaster state and is therefore not of interest to the market.
The shadow bank equity buffer constraint is specified by:
Rsbt + κ
ft (1 + i f
t )B f ,sbt + κe
t (1 + iet)Be,sb
t ≥ (1 + im,sbt )(Msb
t + Msb,xt ),
(4.17)
where 0 < κft , κe
t < 1 specify the percentage of value that is retrieved when
the shadow banks liquidate their assets in the interbank market. It is pos-
sible to read the participation constraint as a “worst case scenario outcome”
which, for the shadow bank, is the realization of a pessimistic signal. In that
case, households liquidate their deposits and shadow bank must sell their
assets in the interbank market to remunerate the households.
In Appendix 4.C we show that for both banks the budget and equity
buffer constraints hold with equality. Hence, it is possible to substitute the
constraints (4.15) and (4.16) in (4.13) to obtain the regulated bank maxim-
ization problem. Likewise, substituting the constraints (4.15) and (4.17) in
(4.13) gives the shadow bank maximization problem. From these maximiz-
ation problems we can induce that as long as the return on a loan is larger
than the costs of funding and insurance, both banks will maximize asset
holdings and minimize the amount of equity funding.
4.3.4 How do banks invest?
Regulated banks pay for deposit insurance costs summarized by the para-
meter χ which is assumed to be a fixed percentage of the total asset return.
Shadow banks circumvent the deposit insurance costs but are limited in
their creation of deposits by the expected liquidation value κιt. Equating the
weighted average costs of funding of regulated banks and shadow banks,
see Appendix 4.D, we can distinguish three different scenarios depending
on the relative weighted average costs of funding of regulated banks versus
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 141PDF page: 141PDF page: 141PDF page: 141
Mortgage loans and shadow banks 131
shadow banks. Proposition 4.1 summarizes the three scenarios:
Proposition 4.1.
a) If κet <
πb1+χie
tand κ
ft < πb
1+χi ft, liquidity in the interbank market for both corporate
loans and mortgages is too low for shadow banks to be able to compete with regulated
banks. Regulated banks specialize in both mortgages and corporate loans.
b) If κet < πb
1+χiet
and πb
1+χi ft< κ
ft , the interbank market for mortgages is suffi-
ciently liquid for shadow banks to overcome the expected liquidation cost disadvant-
age. Regulated banks specialize in corporate loans while shadow banks specialize in
mortgages.
c) If κet > πb
1+χiet
and πb
1+χi ft> κ
ft , the interbank market for corporate loans is suf-
ficiently liquid for shadow banks to overcome the expected liquidation cost disad-
vantage. Regulated banks specialize in mortgages while shadow banks specialize in
corporate loans.
Proof in Appendix 4.D. It is, of course, possible that shadow banks have a
competitive advantage in supplying both mortgage loans and corporate loans.
However, it is not possible for shadow banks to exist without regulated
banks because the viability of the shadow bank business model depends
on the existence of regulated banks that buy shadow bank assets in case of
a liquidation. If shadow banks have lower costs in supplying both corpor-
ate loans and mortgages they specialize in the assets in which they have a
comparative advantage which brings us back to either Proposition 4.1-b or
c.
4.3.5 Expected liquidation value
Whether regulated banks or shadow banks have lower funding costs de-
pends on the liquidation value κιt. We assume that the fundamental liquid-
ation values are affected by the relative depth of the interbank market for
mortgages or corporate loans and by aggregate liquidity risk which depends
on the relative share of deposits that are uninsured, see Stein (2012) and
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 142PDF page: 142PDF page: 142PDF page: 142
132 Chapter 4
Hanson et al. (2015) for a similar approach. Accordingly, the liquidation val-
ues per unit of investment are specified by:
κιt = (1−ω)
(Et|S′=L(πs)
)ϕι
1
(Bι
tBt
)ϕι
2
(Msb
tMt
), (4.18)
where ω = P(d|L)(πb − πd) denotes the deposit insurance costs regulated
banks must pay when they buy shadow bank assets, Et|S′=L(πs) denotes the
fundamental price of the loan when a pessimistic signal occurs and ϕι (·)is a function that specifies the impact of a deeper interbank market for cor-
porate loans and mortgages (Bιt) relative to the entire market (Bt) and more
uninsured deposit creation (Msbt ) relative to total bank deposits (Mt) on the
liquidation value.
The intuition behind (4.18) is as follows. When regulated banks buy
shadow bank loans, regulated banks must insure these loans in the deposit
guarantee system. The term (1−ω) incorporates the insurance costs which
only consist of the probability that a bad state of the world will occur con-
ditional on the occurrence of a pessimistic signal. In Appendix 4.E we show
that when shadow banks diversify their asset portfolio, the liquidation costs
decrease because idiosyncratic risk no longer needs to be insured. This provides
an incentive for shadow banks to diversify their portfolio. In particular, if
the diversification costs are sufficiently low, shadow banks will always com-
pletely diversify their portfolio because a diversified portfolio has a higher
liquidation value.
In case a pessimistic signal about the future state of the world occurs
(S′= L), the expected returns fall from Et(πs)iι
t+1 = 1 to Et|S′=L(πs)iιt+1
< 1. The loan contracts, however, are predetermined at the beginning of
period t under a condition of zero bank profits and therefore the lending
rate equals the banks’ funding costs (the net present value of a loan is zero).
Consequently, the net present value of a bank loan becomes negative when
a pessimistic signal occurs. The reverse holds when a signal about the fu-
ture state of the world is positive. Regulated banks will therefore only buy
shadow bank loans if the price of these loans falls sufficiently to compensate
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 143PDF page: 143PDF page: 143PDF page: 143
Mortgage loans and shadow banks 133
for the loss in value. The fundamental liquidation value after a pessimistic
signal occurs therefore yields Et|S′=L(πs)/Et(πs) < 1 where Et(πs) = 1.
The function ϕι (·) consists of two terms that create a wedge between the
fundamental liquidation value and the value in the interbank market. First,
the depth of the interbank market positively affects the expected liquidation
value as a deeper interbank market makes it easier to find a counterparty,
∂ϕι(·)/∂Bιt > 0. We assume that the relative ease to find a counterparty de-
pends on the size of the market for asset ι (Bιt) relative to the total size of
the market (Bt). Motivated by the fact that individual banks trade in the in-
terbank market, e.g., to construct diversified portfolios or because they face
idiosyncratic liquidity shocks, a larger banking sector in terms of assets res-
ults in more interbank trading. In the limit, when the depth of the market
goes to infinity, no liquidation discount is added and the expected liquida-
tion price of the asset equals its fundamental value: lim(Bιt/Bt)→0ϕι(·) = 1.
Without an interbank market for the asset, the asset can never be sold to
another party and the liquidation price equals zero: lim(Bιt/Bt)→1ϕι(·) = 0.
Second, the creation of shadow bank deposits decreases the expected li-
quidation value, ∂ϕι(·)/∂Msbt < 0. In case a pessimistic signal occurs, more
shadow bank deposits are withdrawn and more shadow bank assets are
sold. Regulated banks buy shadow bank assets, but face aggregate liquidity
risk as households might convert some of their deposits into cash. We as-
sume that the ability of regulated banks to guarantee the safety and liquidity
of their deposits depends on liquidity in the interbank market. The expected
market liquidity therefore depends on the amount of shadow bank deposits
(Msbt ) relative to the total amount of deposits (Mt). In the limit, when all de-
posits are created by shadow banks, the expected liquidation value goes to
zero because all shadow banks sell in case a pessimistic signal occurs and no
regulated bank exists that buys, lim(Msbt /Mt)→∞ϕι(·) = 0. The interbank
market ceases to exist. Without shadow banks, no liquidation discount is
added because no bank ever liquidizes its assets, lim(Msbt /Mt)→0ϕι(·) = 1.
Brunnermeier and Pedersen (2009), Stein (2012) and others showed that
the creation of uninsured deposits is associated with a negative externality.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 144PDF page: 144PDF page: 144PDF page: 144
134 Chapter 4
Banks do not take into account that selling an asset affects the price of the
asset which leads to further losses. Consequently, the initial decrease in the
liquidation value is amplified and the ex-ante expected liquidation value
differs from the ex-post realized liquidation value. In this chapter we take
the negative correlation between more shadow bank deposits and the ex-
pected liquidation value as given ∂κι/∂Msbt < 0. Instead, we analyze why
this negative externality is not priced by the market.
4.3.6 Externalities
At the aggregate level the depth of the interbank market and the amount of
shadow bank deposits affect the expected liquidation value of shadow bank
assets. If, however, shadow banks take the expected liquidation discount as
given and do not include the incremental impact they have on its value,
two externalities emerge. In this case, shadow banks maximize their profits
(4.13) subject to the budget (4.15) and bank equity constraints (4.17). The
FOCs w.r.t. loans and deposits yield:
iιt = iq,sb
t − µsbt κι
t(1 + iιt), (4.19)
im,sbt = iq,sb
t − µsbt (1 + im,sb
t ), (4.20)
where µsbt denotes the shadow value of the bank equity buffer constraint.
We use that the shadow value of the budget constraint equals the costs of
bank equity, see Appendix 4.C. In contrast, maximizing shadow bank profits
w.r.t. loans and deposits taking into account the incremental impact shadow
banks have on the liquidation parameter yields:
iιt =iq,sb
t − µsbt (1 + iι
t)
[kι
t +∂kι
t∂Bι
tBι
t
], (4.21)
im,sbt =iq,sb
t − µsbt
((1 + im,sb
t )−∑ι
∂kι
t∂Ms
t(1 + iι
t)Bιt
). (4.22)
The impact of ignoring the incremental impact of shadow banks on the li-
quidation parameter is twofold. First, an increase in credit supply for mort-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 145PDF page: 145PDF page: 145PDF page: 145
Mortgage loans and shadow banks 135
gages or corporate loans by shadow banks increases the depth of the market
for mortgages and corporate loans. Consequently, the shadow bank equity
buffer constraint relaxes because µsbt (1 + iι
t)∂κιt/∂Bι
t > 0. Hence, if shadow
banks neglect the incremental impact they have on the depth of the interb-
ank market, they underestimate their marginal costs. Consequently, lending
rates are higher and credit supply and shadow bank leverage are lower than
their social optimal levels.
Second, from a funding perspective, if shadow banks issue deposits,
market liquidity decrease. The shadow bank equity buffer constraint always
binds and accordingly shadow bank deposit creation is at its maximum at-
tainable level. If a shadow bank changes its capital structure by financing
a larger share of its assets with deposits, it realizes a private benefit in the
form of lower financing costs. However, the shadow bank also decreases the
expected liquidation value of other shadow banks represented by a decrease
in κιt because more shadow banks assets are sold when a pessimistic signal
occurs, ∂κιt/∂Ms
t < 0. The equity buffer constraint of all other shadow banks
tightens. Consequently, the deposit rate, shadow bank deposit creation and
shadow bank leverage are above their socially optimal level. The following
proposition summarizes the discussion:
Proposition 4.2.
Let B∗ and M∗ denote the social optimal amounts of shadow bank loans and deposits
and let B′
and M′
denote the optimal amount of shadow banks loans and deposits
from the perspective of the shadow banks. Shadow banks supply too much credit,
B∗ < B′, and create too many deposits, M∗ < M
′, if |∂κι
t/∂Bιt| < |∂κι
t/∂Mst |.
4.3.7 Closure
The real side of the model is closed by imposing the goods market equilib-
rium. Total production is equal to consumption, investment and lump-sum
central bank consumption (which is equal to the deposit premium paid by
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 146PDF page: 146PDF page: 146PDF page: 146
136 Chapter 4
regulated banks if not paid out):
Yt = Cpt + Ci
t + It + Dt. (4.23)
We assume foreign deposit demand takes the same functional form as
the domestic deposit demand equation for σm = −1. All endogenous vari-
ables affecting deposit demand (the foreign deposit, lending and bank equity
rate, consumption and housing), except for the domestic deposit rate, are
lumped in a fixed term ϑ. The domestic interest rate provides feedback as an
increase in foreign deposit demand reduces the domestic interest rate which
attenuates the increase in foreign deposit demand. The functional form is
represented by:
Mrb,xt = (εm
t − 1)ϑ log(1 + im,it ), (4.24)
where εmt denotes a foreign deposit demand shock, i.e., a savings glut if
εmt > 1, εm
t = exp(ηmt ), where ηm
t = ρmηmt−1 + εm
t and εmt is an i.i.d. error term
∼ (0, σm).
Housing supply is fixed at an arbitrarily level H:
Hst = H. (4.25)
Prices are perfectly flexible and accordingly there is no role for conven-
tional monetary policy to attenuate macroeconomic fluctuations by setting
the policy rate. However, as Goodhart (1988) argues, the original motivation
for creating most central banks was not maintaining price stability but, as in
this chapter, to provide financial stability. Here the central bank regulates
the banks and fulfills the lender of last resort function when the banking
sector has insufficient funding. Aggregate liquidity can only fall when ag-
gregate demand for cash Mcbt increases. In this case, the banking sector faces
a funding shortfall and needs to borrow from the central bank. Bank bor-
rowing from the central bank is denoted by Bcbt . The central bank balance
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 147PDF page: 147PDF page: 147PDF page: 147
Mortgage loans and shadow banks 137
sheet is represented by:
Bxt + Bcb
t = Rt + Mcbt . (4.26)
In normal times both non-interest paying deposits at the central bank Rt and
household holdings of cash Mcbt are zero such that borrowing from the cent-
ral bank by banks, Bcbt , is also zero. Moreover, Bx
t denotes borrowing from
the central bank by non-domestic banks (denoted in the domestic currency).
Borrowing by foreign banks is zero in equilibrium and increases when the
aggregate foreign banking sector has insufficient funding.
The intuition behind the transmission of the shock εmt is as follows. For-
eign households withdraw their deposits at foreign banks and deposit these
deposits at domestic banks. Foreign demand for domestic deposits, Mrb,xt ,
increases. As foreign banks have a funding shortfall, they can either bor-
row from domestic banks or the central bank. We assume that the interbank
market does not provide funding for these foreign funding shortfalls. For-
eign banks therefore lend from the central bank, while domestic banks de-
posit their excess funds at the central bank. Both Bxt and Rt increase. Central
bank deposits do not pay any interest, while bank deposits do pay interest.
Therefore the domestic banking sector decreases the deposit rate to reduce
demand for its deposits. The decline in the deposit rate decreases domestic
and foreign deposit demand and it increases borrowing by impatient house-
holds.
4.3.8 Calibration
Table 4.2 specifies the parameter settings. Patient and impatient consumers
have a slightly different discount factor to ensure that impatient house-
holds borrow and that patient households save. The coefficient that determ-
ines the relative risk aversion of households—the substitution elasticity of
consumption—σc is set at 1 ensuring log-utility and the inverse of the elasti-
city of work effort with respect to the wage rate σl = 2 are set at conventional
values, see Christiano et al. (2005). The substitution elasticity of deposits σm
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 148PDF page: 148PDF page: 148PDF page: 148
138 Chapter 4
Table 4.2. Calibrated parameters
Parameters Description Value
βp Discount factor patient households 0.990βi Discount factor impatient households 0.989θ Share of patient households 0.500η Share of housing in the consumption bundle 0.250γ
pl Weight of leisure in utility function patient households 0.627
γil Weight of leisure in utility function impatient households 1.920
γm Weight of deposits in utility function 0.040α Share of labor in the production function 0.667δ Capital depreciation rate 0.025φ Capital adjustment costs 2.500σc Substitution elasticity consumption 1.000σm Substitution elasticity deposits holdings 1.000σl Substitution elasticity leisure 2.000H Housing supply 1.000ϑ Exogenous foreign deposit demand 10.000ϕι
1 Market depth feedback parameter 0.200ϕι
2 Shadow bank deposits feedback parameter 0.200πg Productivity good state 1.100πb Productivity bad state 0.850πd Productivity disaster state 0.100P(u) Probability good signal 0.750P(u|H) Probability good state if good signal 0.950P(u|L) Probability good state if bad signal 0.800P(d|L) Probability disaster state if bad signal 0.100ρm Persistence parameter increase in deposits shock 0.900µm Mean increase in deposits shock 0.000σm S.D. savings glut shock 1.000
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 149PDF page: 149PDF page: 149PDF page: 149
Mortgage loans and shadow banks 139
is set at 1 which also ensures log-utility, but more importantly, a direct link
with consumption. Moreover, we set the weight of housing in the consump-
tion bundle (1− η) equal to 0.25 which entails that housing is 1/4 of total
consumption expenses. The weights of labor and deposits γl and γm in the
utility function are used to normalize labor supply in steady state of both
representative households to unity and bank leverage in steady state equal
to Q/(Q + M) = 20%.
The physical capital adjustment cost parameter φ is set at 2.5, close to the
value estimated by Christiano et al. (2005). The physical capital depreciation
rate δ and the share of labor in the production function α take their standard
values of 0.025 per quarter and 2/3, respectively. The loan-to-value paramet-
ers ζ f and ζe are set equal to unity in the benchmark case. Housing supply
H is fixed to unity. The supply elasticity of foreign deposit demand with
respect to the domestic deposit rate ϑ equals its domestic value in steady
state.
The probabilities that a state of the world occurs and the corresponding
productivities are set such that the ex-ante expected value of Et(πs) equals
1. The expected productivity after the bad signal is equal to approximately
90% and the DGS costs are set to approximately 1% of the bank balance
sheet.
4.4 Results
4.4.1 Growth of mortgage loans
Figure 4.9 shows the effects of an exogenous increase in funding on the ag-
gregate bank balance sheet. In this section, we set shadow banks’ costs of
funding equal to regulated banks’ costs of funding by setting the liquida-
tion value equal to a constant κι = πb/(1 + χ)iι∗. The results for the three
cases discussed in Proposition 4.1 are almost identical because the real side
of the economy is similar. The banking structure, however, is slightly differ-
ent which might give rise to a small difference. Specifically, regulated banks
pay DGS costs which reduces the amount of funding that could be used
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 150PDF page: 150PDF page: 150PDF page: 150
140 Chapter 4
Figure 4.9. An increase in bank deposits and the real economy
Notes: Impulse responses following an increase in bank deposits, that is, a positive shock toεm
t while the expected asset liquidation values (κet and κ
ft ) remain constant. The increase in
deposits equals 3% of total domestic deposits Mt. Horizontal axis shows quarters. Verticalaxis shows deviations from steady state.
for lending compared to the uninsured shadow banking sector. Second, the
bank equity constraints of the two banking types are slightly different. This
difference marginally impacts their financial structure and it therefore af-
fects the amplification of fluctuations.
The increase in domestic bank deposits raises banks’ holdings of cent-
ral bank deposits contemporaneously. Although domestic banks could also
lend to the foreign banks who face a shortfall, here we assume that domestic
banks do not want to lend to foreign banks because, e.g., they consider for-
eign banks too risky. Banks lower the deposit rate because aggregate bank
deposits increase, while the resulting central bank deposits do not generate
any extra income. Both the mortgage rate and the corporate loan rate follow
the reduction in bank funding costs because banks compete which eventu-
ally equalizes their lending rate to their weighted average costs of funding.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 151PDF page: 151PDF page: 151PDF page: 151
Mortgage loans and shadow banks 141
Figure 4.10. An increase in bank deposits and the share of mortgages (%balance sheet)
Notes: Impulse responses following an increase in bank deposits, that is, a positive shock toεm
t while the expected asset liquidation values (κet and κ
ft ) remain constant. The increase in
deposits equals 3% of total domestic deposits Mt. Horizontal axis shows quarters. Verticalaxis shows deviations from steady state. Mortgage loans as % of the aggregate bank balance
sheet are calculated as: B ft
B ft +Be
t+Dt. The mortgage-to-income ratio is calculated as: B f
tWt Li
t+rkt Kt−1
.
Hence, arbitrage and competition ensure that both lending rates fall and
borrowing increases.
Patient and impatient households hold different asset portfolios. Patient
households hold bank deposits, bank equity and houses while impatient
households hold bank deposits, houses and physical capital. The return on
these assets changes which provides an incentive for both types of house-
holds to rebalance their portfolio. The inflow of deposits raises bank lever-
age and therefore the required return on bank equity increases. As the lever-
age constraint is binding, banks induce patient households to rebalance their
portfolio from deposits and housing towards bank equity. This implies that
the return on bank equity increases relative to the expected return on hous-
ing and deposits. Patient households therefore reduce savings, sell part of
their housing stock to the impatient households and convert the proceeds
and some of their deposits into bank equity.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 152PDF page: 152PDF page: 152PDF page: 152
142 Chapter 4
Impatient households do not own—and do not wish to own—bank equity
and are therefore unaffected by changes in the return on bank equity. That
is, the return on bank equity is still below the required return of impatient
households to due their high discount factor. For banks the costs of fund-
ing decreases because the decline in the interest rate on deposits is larger
than the increase in the required return on bank equity. As the weighted
average costs of funding falls, banks increase their lending by lowering the
lending rate on mortgages and corporate loans. When lending rates decline,
impatient households prefer to increase their indebtedness and buy more
houses and physical capital. Impatient households’ demand for houses and
physical capital given limited supply of both assets drives house prices and
physical capital prices up.
The increase in mortgage lending is stronger than the increase in corpor-
ate lending. Figure 4.10 shows that the total amount of mortgage loans as
percentage of the aggregate bank balance sheet increases. To understand this
outcome, Figure 4.11 plots the expected return on physical capital Etrkt+1,
housing Etrht+1 and bank equity Etre
t+1. As lending rates fall, housing
demand by impatient households increases which drives up house prices.
This same mechanism is at work for physical capital: the increase in demand
for physical capital drives up the price of physical capital and thereby the ex-
pected return on physical capital. However, when patient households obtain
funding to increase the production of physical capital, the return on phys-
ical capital falls. For housing demand, no offsetting supply effect is present.
The house price increase relaxes the collateral constraint for mortgage loans
more than the collateral constraint for corporate loans. Consequently, the
supply of mortgage loans can increase by more than the supply of corporate
loans and bank investment is reallocated towards mortgages.
The aggregate housing stock is fixed, but from the perspective of the im-
patient household housing supply increases. This housing supply effect is,
in the model, induced by patient households who sell part of their housing
stock to impatient households. Patient households are willing to do so be-
cause the expected return on bank equity increases even more. These results
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 153PDF page: 153PDF page: 153PDF page: 153
Mortgage loans and shadow banks 143
Figure 4.11. Expected return on physical capital, housing and bank equityfollowing an increase in bank deposits
Notes: Impulse responses following an increase in bank deposits, that is, a positive shock toεm
t while the expected asset liquidation values (κet and κ
ft ) remain constant. The increase in
deposits equals 3% of total domestic deposits Mt. Horizontal axis shows quarters. Verticalaxis shows deviations from steady state. The expected return to physical capital Etrk
t+1 isrepresented by the circled (black) line, the expected return to housing Etrh
t+1 is representedby the barbed (blue) line and the expected return to bank equity Etre
t+1 is given by the bar-striped (red) line.
imply that both a reduction in the supply elasticity of housing and a de-
crease in the share of households buying a house with their own wealth, in-
creases house price fluctuations and therefore mortgage supply fluctuations.
4.4.2 Shadow bank growth
In this section we link the relative growth of mortgage loans to the growth
rate of the shadow banking sector. In Section 4.4.1 the liquidation value was
set equal to a constant. In the remainder of this chapter, the liquidation value
is endogenously determined by Equation (4.18). Accordingly, the relative
growth of mortgage loans following the decrease in interest rates described
in Figure 4.9 deepens the interbank market for mortgage loans. The expected
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 154PDF page: 154PDF page: 154PDF page: 154
144 Chapter 4
Figure 4.12. Expected asset liquidation value following an inflow of deposits
Notes: Impulse responses following an increase in bank deposits, that is, a pos-itive shock to εm
t . Horizontal axis quarters. Vertical axis shows deviations from
steady state. The functional form for the asset liquidation value is: ϕι(
Bιt
Bt, Msb
tMt
)≡(
1 + ϕι1 log
Bι
t
Bιt+Bι′
t
)(− log
ϕι
2 Msbt
Mt
). The barbed (blue) line represents the expected li-
quidation value for mortgage loans (κ ft ) without aggregate liquidity risk (ϕ
f1 = 1 and the
second term in round brackets on the right hand side equals 1). The circled (black) linesrepresented the expected liquidation value for mortgage loans with high aggregate liquidityrisk (ϕ
f1 = 1 and ϕ
f2 = 6). The bar-striped (red) line shows the expected liquidation value for
corporate loans (κet ) without aggregate liquidity risk (ϕe
1 = 1 and the second term in roundbrackets on the right hand side equals 1).
liquidation value κft increases which improves the comparative advantage
of shadow banks vis-a-vis regulated banks in supplying mortgage loans.
Positive feedback between the size of the shadow banking sector and
the asset liquidation value, ∂κ f /∂Bι > 0, allows the shadow banking sector
to grow. Specifically, more investment in mortgage loans by shadow banks
increases the size of and thereby liquidity in the interbank market for mort-
gage loans. This increases the expected asset liquidation value and therefore
the amount of uninsured deposits shadow banks can create by creating new
loans. In equilibrium, this positive feedback between the size of the interb-
ank market and the expected liquidation value internalizes the externality:
∂κι/∂Bι → 0.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 155PDF page: 155PDF page: 155PDF page: 155
Mortgage loans and shadow banks 145
The barbed line in Figure 4.12 shows that the market for mortgage loans
becomes more liquid—the expected liquidation value for shadow bank as-
sets increases—when banks receive an inflow of deposits. Even though shadow
banks do not take the effect of their own behavior on the assets’ liquidation
value into account, liquidity in the market is improving. The comparative
advantage of shadow banks with respect to supplying mortgages increases.
The dotted line in Figure 4.12 shows how liquidity in the interbank mar-
ket for corporate loans is decreasing, compared to the interbank market for
mortgages, because the market for corporate loans is shrinking. Hence, an
exogenous inflow of deposits in banks increases liquidity in the market for
mortgage loans. This in turn fosters the comparative advantage of shadow
banks over regulated banks.
4.4.3 Liquidity risk and the lender of last resort
In the model, liquidity risk in the banking sector depends on the size of the
shadow banking sector. When a pessimistic signal occurs, two things can
happen. First, if the shadow banking sector is relatively small compared to
the regulated banking sector,(
Msbt /Mt ≈ 0
), the interbank market contin-
ues to function efficiently as only a small share of the banking sector sells
assets. In this case the regulated banking sector buys—or lends the funding
shortfall to—the shadow banking sector. Regulated bank deposits increase
and aggregate liquidity is unaffected because households do not increase
their holdings of cash: Mcbt = 0 and ∆Mrb
t = ∆Msbt .
Second, if the shadow banking sector is relatively large(
Msbt /Mt ≈ 1
),
a large share of the interbank market starts to sell assets. Without sufficient
activity in the interbank market regulated banks cannot diversify their as-
sets and/or insure idiosyncratic liquidity risks. It becomes increasingly diffi-
cult for regulated banks to guarantee the safety and liquidity of their depos-
its. Regulated bank deposits might therefore be converted into cash: Mcbt > 0
and ∆Msbt > ∆Mrb
t . In this case, aggregate liquidity decreases and regulated
banks need additional funding from the central bank.
Regulated banks can borrow from the central bank or reduce their lend-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 156PDF page: 156PDF page: 156PDF page: 156
146 Chapter 4
ing to the central bank to compensate for the decrease in depository fund-
ing. When households increase their holdings of cash, the liability side of the
central bank balance sheet increases, see Equation (4.26). The central bank
balance sheet implies that either central bank deposits Rt should decrease
or bank borrowing from the central bank Bcbt should increase. However, for
regulated banks to have a level of central bank deposits Rt > 0, they must
have borrowed from the central bank in the first place Bcbt = Rt. In this case,
∆Mcbt = −∆Rt. In steady state banks do not borrow from the central bank to
hold (costly) central bank deposits in excess of any requirement. Therefore
Rt cannot decrease to compensate for the deposit outflow. Alternatively, bor-
rowing from the central bank must increase ∆Bcbt = ∆Mcb
t . In other words,
by providing cash to households via the banking sector, the central bank
must also provide liquidity insurance to banks.
In practice, the central bank lends to regulated banks if they have ad-
equate collateral. There is no need to internalize the increase in liquidity risk
if the market expects that the central bank will always lend to the regulated
banking sector because they have sufficient adequate collateral. The central
bank is the lender of last resort and can provide regulated banks additional
funding equal to: ∆Bcbt = ∆Msb
t − ∆Mrbt . Buying shadow bank assets poses
no threat to the regulated banks’ future funding position, if they can always
borrow from the central bank. Regulated bank deposits remain safe and li-
quid and depositors are unlikely to convert their regulated bank deposits
into cash. Hence, the risk associated with the creation of uninsured shadow
bank deposits, ∂κιt/∂Ms
t , depends on whether the central bank is likely to
lend to regulated banks.
A growing shadow banking sector increases the regulated banks’ reli-
ance on liquidity support by the central bank. Yet, regulated banks do not
pay a fee for this liquidity insurance. Although it becomes more likely that
the central bank must provide liquidity support when the shadow banking
sector grows, the costs of liquidity support by the central bank are unaf-
fected. In contrast, borrowing from the central bank is even likely to become
less expensive if the central bank is expected to lower the policy rate in re-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 157PDF page: 157PDF page: 157PDF page: 157
Mortgage loans and shadow banks 147
sponse to a financial crisis. Also, as has been the case during the global finan-
cial crisis, the central bank might accept a broader set of assets as collateral
when all banks have a funding shortness. Consequently, shadow banks cre-
ate too many uninsured deposits compared to the social optimum, i.e., they
expose the banking system to excessive liquidity risk, because liquidity in-
surance by the central bank is not priced.
In this case, the increase in market liquidity as a consequence of a deeper
interbank market outweighs the increase in funding liquidity risk as the lat-
ter effect is not fully priced. The barbed line in Figure 4.12 shows that the
expected asset liquidation value increases and the shadow bank equity con-
straint (4.17) is relaxed when shadow bank lending increases because the
actual increase in liquidity risk is not internalized. The circled line in Figure
4.12 shows how liquidity in the market for mortgages actually decreases
when the increase in liquidity risk is internalized. The difference between
the barbed and the circled line represents the value of the liquidity insur-
ance provided by the central bank.
4.5 Policy options
4.5.1 Loan-to-value constraints
A reallocation of bank investment towards mortgage loans can adversely
impact economic growth and financial stability. Such a reallocation does not
depend on the financial structure of the banking sector and is not associ-
ated with an externality. In fact, the reallocation is determined by limited
housing supply and the collateral constraints for both mortgage loans and
corporate loans work, if anything, in the opposite direction limiting credit
supply. However, an increase in household debt relative to income makes it
harder to repay the debt. Indeed, Figure 4.10 shows that the loan is financed
to a larger extent by an increase in collateral value and not by an increase
in household income. Although the model excludes actual Ponzi schemes
by construction, in practice it might be harder to distinguish an increase in
mortgage debt supported by fundamentals from an increase in household
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 158PDF page: 158PDF page: 158PDF page: 158
148 Chapter 4
Figure 4.13. An increase in bank deposits and restrictions on admissibleloan-to-value ratios
Notes: Impulse responses following an increase in bank deposits, that is, a positive shock toεm
t . Restrictions on admissible loan-to-value ratios, circled (black) line LTV is 100% (ζ f = 1)and barbed (red) line LTV is 80% (ζ f = 0.8). Horizontal axis quarters. Vertical axis showsdeviations from steady state.
debt supported by Ponzi finance (Minsky, 1986).
Prudential authorities in some countries have recently responded to the
increase in household debt relative to household income and/or house value
by introducing restrictions on admissible loan-to-value (LTV) and loan-to-
income ratios. These restrictions should create a precautionary buffer against
house price fluctuations. From Equations (4.10) and (4.11) and the analysis
in Appendix 4.B, we conclude that the steady-state loan-to-value-ratio falls
when these constraints tighten. A lower LTV ratio provides a buffer against
future house price fluctuations.
Figure 4.13 reveals two additional benefits of restrictions on admissible
LTV ratios. First, stricter constraints on admissible LTV ratios limit house
price and thereby mortgage supply fluctuations when shocks hit the eco-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 159PDF page: 159PDF page: 159PDF page: 159
Mortgage loans and shadow banks 149
nomy. Although the restrictions by themselves are not sufficient to prevent
the reallocation of investment and thereby shadow banking growth, they do
attenuate bank investment in mortgage loans and thereby shadow banking
growth. Hence, restrictions on admissible LTV ratios provide a larger buffer
against house price fluctuations, and simultaneously attenuate fluctuations
in house prices and mortgage loans. Second, restrictions on admissible LTV
ratios reallocate bank investment from mortgage loans to corporate loans.
This effect is not only present in the steady state but also when credit sup-
ply expands. As less lending are re-allocated towards houses, the fall in in-
terest rates following the inflow of deposits has a stronger positive effect
on investment. Hence, restrictions on admissible LTV ratios can reallocate
investment to physical capital while it limits household mortgage debt and
house price growth.
4.5.2 Interest on cash
The increase in liquidity risk due to the creation of uninsured shadow bank
deposits is, if not internalized by the banks, a pecuniary externality and
socially excessive. Pecuniary externalities violate the first welfare theorem
when the liquidation value affects not only the bank’s budget constraint, but
also its collateral constraint. As the liquidation price is present in the bank
equity constraint, the first welfare theorem is violated. The growth of mort-
gage debt increases regulatory arbitrage, but the banking sector does not
price the increase in liquidity risk associated with more uninsured shadow
bank deposits. The shadow banking sector creates too many uninsured de-
posits compared to the social optimum which leaves the financial system
excessively vulnerable to a liquidity crisis.
The key externality driving a wedge between private and social optimal
values is the negligence of an increase in liquidity risk when shadow banks
create uninsured deposits. As the creation of insured deposits creates liquid-
ity risk, the central bank provides liquidity insurance. However, since li-
quidity insurance is not priced, banks issue too many deposits rather than
equity leaving the financial sector vulnerable to liquidity crises. A deposit
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 160PDF page: 160PDF page: 160PDF page: 160
150 Chapter 4
insurance scheme reallocates the risk of creating uninsured deposits outside
the regulated banking sector towards the shadow banking sector. This how-
ever, does not solve the financial systems’ reliance on liquidity insurance
provided by the central bank.
Central banks can realign private and social interests by regulating the
total amount of (uninsured) bank deposits. One possibility is a Pigouvian
tax as suggested by Stein (2012). In his proposal the central bank regulates
the total amount of private uninsured deposit creation by introducing a flex-
ible cap-and-trade system in which banks are granted tradable permits to
create uninsured deposits. The price of these permits reveals information
about the banking sectors’ investment opportunities. The regulator can ad-
just the amount of permits in the system in accordance with its objectives
when prices change.
Although the cap-and-trade system proposed by Stein (2012) is an effect-
ive instrument to regulate the creation of uninsured deposits by regulated
banks, it has adverse consequences on shadow banks and the real economy.
For one thing, the system does not allow the regulator to observe the op-
timal level of permits. When the price of permits increases, the banking sec-
tor might have better investment opportunities, but it does not signal how
much investment is optimal. Moreover, restricting the creation of uninsured
deposits increases regulatory arbitrage as deposits become more scarce. Ac-
cordingly, shadow banks, who are not regulated, have a larger incentive to
create uninsured deposits while any liquidity risk remains insured by the
central bank. The cap-and-trade system therefore does not eliminate the key
externality, but relocates risks outside the regulated banking sector.
Here, we suggest a more direct approach: the central bank pays interest
on cash. Thereby the central bank eliminates the incentive for both regulated
banks and shadow banks to create uninsured deposits. The basic idea is
that paying interest on cash raises the opportunity costs for households to
deposit their savings in regulated and shadow banks. If the interest rate on
cash is set higher than the interest rate on bank deposits, banks must raise
the deposit rate to attract depository funding thereby reducing the incentive
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 161PDF page: 161PDF page: 161PDF page: 161
Mortgage loans and shadow banks 151
to finance themselves with deposits rather than equity.
Deposits can be used for consumption while equity cannot. From the
patient households’ FOC with respect to deposits and after substituting the
FOC with respect to bank equity, we can induce that households are willing
to pay a premium for deposits relative to equity:
(Mjt)
σm=
γm
λpt
(1 + iq
t
iqt − im
t
). (4.27)
From (4.27) we find that if deposit demand is positive, the rate on deposits
is lower than the required return on equity: iqt > im
t . Consequently, banks
have an incentive to finance themselves with deposits rather than equity.
Cash can also be used for consumption, but households limit their hold-
ings of it because cash does not pay interest. We therefore propose that the
central bank pays interest on cash to compete with the banking sector in
the creation of liquid, safe claims. In the initial case where the central bank
pays no interest im,cbt = 0, households will only hold cash for consumption
purposes when no alternative is available. If the central bank pays interest
on cash which is higher than the interest rate on bank deposits im,cbt > im,s
t ,
depositors would like to hold only cash and no bank deposits. In this case,
both types of banks must raise the deposit rate to retain their deposits, fin-
ance themselves fully with equity or borrow from the central bank for the
same rate: im,cbt . Consequently, when im,cb
t = iqt banks become indifferent
between debt and equity finance as both have exactly the same costs. That
is, the Modigliani and Miller (1958) irrelevance proposition is no longer vi-
olated.
In practice the central bank can set the interest rate on central bank de-
posits equal to the interest rate banks receive on their deposits at the cent-
ral bank. At this interest rate the externality associated with the creation of
uninsured deposits is eliminated. In the model presented in this chapter,
the (risk-free) required return on bank equity is equal to the interest rate
banks receive on their deposits at the central bank. The deposit rate trades
at a discount of this rate as consumers are willing to accept a lower interest
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 162PDF page: 162PDF page: 162PDF page: 162
152 Chapter 4
rate for liquid and safe deposits (the literature refers to this difference as
the money premium). The difference between the policy rate and the de-
posit rate stimulates banks to finance themselves with deposits rather than
equity. Hence, if the central bank offer households a savings account that
pays an interest rate equal to the interest rate banks receive on their deposits
at the central bank, this difference disappears. Depositors become indiffer-
ent between central bank deposits and bank deposits while banks become
indifferent between deposit and equity finance.
4.6 Conclusion
In this chapter we showed how an exogenous decline in real interest rates
caused by an inflow of deposits could explain both the reallocation of bank
investment from corporate loans to mortgages and the growth of the shadow
banking sector relative to the regulated banking sector. Specifically, inelastic
housing supply relative to the supply of physical capital causes house prices
to rise which relaxes the collateral constraint for mortgage debt and induces
the reallocation, in relative terms, towards mortgage loans. Positive feed-
back between the depth of the interbank market for mortgage loans and the
liquidation value of shadow bank assets increases shadow banks’ compar-
ative advantage over regulated banks in supplying mortgages.
In the model we showed that the growing shadow banking sector leaves
the banking sector excessively vulnerable to financial crises. When shadow
banks grow and they finance their loans with uninsured deposits, liquid-
ity risk increases. The shadow banking sector relies indirectly on liquid-
ity insurance provided by the central bank. As shadow banks create more
uninsured deposits, the banking system’s reliance on liquidity support by
the central bank increases. However, liquidity insurance is not priced in
the market. The increase in liquidity risk is therefore not fully incorpor-
ated in the expected liquidation price of shadow bank assets. Consequently,
shadow banks issue too many uninsured deposits relative to the social op-
timum.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 163PDF page: 163PDF page: 163PDF page: 163
Mortgage loans and shadow banks 153
Macroprudential regulation—such as restrictions on admissible loan-to-
value ratios—can offer a first line of defense to preserve financial stability
because these constraints re-allocate bank investment back towards corpor-
ate loans. In addition, central banks can remove the incentive for banks to
finance themselves with deposits rather than equity by paying interest on
cash. Thereby the central bank raises the households’ opportunity costs of
holding bank deposits. Banks must increase the deposit rate which reduces
the incentive for banks to finance themselves with deposits rather than bank
equity. This eliminates a key externality and leaves the economy less vulner-
able to liquidity risk.
4.A Variable names and definitions
Table 4.A.1. Variable names and definitions
Notation Source Definition
Shadow bank growth ∆ ln Ysbt Flow of funds ∆ log of other financial inter-
accounts FRB US mediaries except insurance companiesand pension funds; total financial assets
Regulated bank growth ∆ ln Yrbt Flow of funds ∆ log of domestic financial
accounts FRB US sectors; total financial assetsGrowth of mortgage loans ∆ ln Mtotal
t Flow of funds ∆ log all sectors;accounts FRB US total mortgages; asset
Growth of mortgage loans ∆ ln Mpdit Flow of funds ∆ log private depository
(only depository institutions) accounts FRB US institutions; total mortgages; assetGrowth of corporate loans ∆ ln Ctotal
t Flow of funds ∆ log nonfinancial corporateaccounts FRB US business; loans; liability
MBS and ABS markets growth ∆ ln Ztotalt Flow of funds ∆ log of total GSE
accounts FRB US ABS and private MBS marketsInflation (CPI) - Federal Reserve Consumer Price Index for All Urban
Bank of St. Louis Consumers: All Items, Percent Changefrom Year Ago, Quarterly, Seasonally Adj.
Federal Funds Rate - Federal Reserve Effective Federal Funds Rate, Percent,Bank of St. Louis Quarterly, Not Seasonally Adj.
GDP growth - Federal Reserve Real Gross Domestic Product, PercentBank of St. Louis Change from Preceding Period, Quarterly,
Seasonally Adjusted Annual Rate
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 164PDF page: 164PDF page: 164PDF page: 164
154 Chapter 4
Table 4.A.2. Descriptive statistics
Mean Median Maximum Minimum Std. Dev. Jarque-Bera Obs.
Shadow bank growth 10.699 11.655 29.978 -9.868 6.688 16.839 248Regulated bank growth 6.978 7.231 14.416 -6.480 3.484 19.192 248Growth of mortgage loans 7.784 8.760 14.423 -4.142 4.337 30.938 248Growth of mortgage loans 7.148 7.844 16.808 -6.703 5.328 16.124 248(private depository institutions)Growth of corporate loans 6.716 8.408 21.250 -22.023 7.316 129.442 248MBS and ABS markets growth 12.594 12.735 39.739 -17.791 9.561 1.072 248Inflation (CPI) 3.656 3.032 14.426 -1.607 2.817 179.824 248Federal Funds Rate 4.949 4.750 17.780 0.070 3.593 47.745 248GDP growth 3.059 3.050 16.500 -10.000 3.550 27.011 248
Figure 4.A.1. VAR estimation results of Equation (4.2) for shadow banks dif-ferent ordering
Corporate loan to shadow bankCorporate loan to mortgage loan Corporate loan to corporate loan
Note: time (year) on the horizontal axis, percentage point devations on the vertical axis.
Shadow bank to shadow bankShadow bank to mortgage loan Shadow bank to corporate loans
Mortgage loan to shadow bankMortgage loan to mortgage loan Mortgage loan to corporate loan
-2
-1
0
1
2
2 4 6 8 10
-2
-1
0
1
2
2 4 6 8 10
-2
-1
0
1
2
2 4 6 8 10
-2
0
2
4
2 4 6 8 10
-2
0
2
4
2 4 6 8 10
-2
0
2
4
2 4 6 8 10
-2
0
2
4
2 4 6 8 10
-2
0
2
4
2 4 6 8 10
-2
0
2
4
2 4 6 8 10
Notes: Ordering Xt = [Mtotalt , Ctotal
t , Ybt , Breal
t , Ztotalt ]. Estimation period 1955Q3–2017Q2,
time (quarters) horizontal axis, percentage point deviation on the vertical axis. The titlesexplain the response of one of the variables included to a shock of any of the variables in-cluded. Dotted red lines denote confidence intervals at 95% significance level.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 165PDF page: 165PDF page: 165PDF page: 165
Mortgage loans and shadow banks 155
4.B Household and firm problem
Patient household problem
Patient household utility function:
U jt = Et
∞
∑t=0
(βj)tεjt
(((Cj
t)η(H j
t)1−η)1−σc
1− σc + γm (Mjt)
1−σm
1− σm − γl (Ljt)
1+σl
1 + σl
).
(4.B.1)
Patient household budget constraint:
Mpt + Qt + qh
t (Hpt − Hp
t−1) =(1 + imt−1)Mp
t−1 + (1 + iqt−1)Qt−1 + rh
t Hpt−1
− Cpt + Ltwt + θ(Πtb
t + Πpt ). (4.B.2)
Patient household Lagrangian:
Lp =E0
∞
∑t=0
(βp)t εpt
(((Cp
t )η(Hp
t )1−η)1−σc
1− σc + γm (Mpt )
1−σm
1− σm − γl (Lpt )
1+σl
1 + σl
)
+ λpt
[(1 + im
t−1)Mpt−1 + (1 + iq
t−1)Qt−1 + rht Hp
t−1 + Ltwt
+ θ(Πtbt + Πp
t )− Cpt − It −Mp
t −Qt − qht (Hp
t − Hpt−1)
], (4.B.3)
where λpt is the Lagrangian multiplier associated with the patient household
budget constraint. The FOCs conditions w.r.t. Cpt , Lp
t , Hpt , Mp
t , Qt are given
by:
ηCp(η−1)t (Cpη
t Hp(1−η)t )−σc
= λpt (4.B.4)
εpt γl(Lp
t )σl= λ
pt wt (4.B.5)
εpt (1− η)Hp(−η)
t (Cpηt Hp(1−η)
t )−σc − λpt qh
t + λpt+1βp(qh
t+1 + rht+1) = 0
(4.B.6)
εpt γm(Mj,ι
t )−σm+ λ
pt+1βp(1 + im
t ) = λpt (4.B.7)
λpt+1βp(1 + iq
t ) = λpt . (4.B.8)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 166PDF page: 166PDF page: 166PDF page: 166
156 Chapter 4
Rewriting the FOCs, substituting out the Lagrangian multipliers λpt and
setting σc = 1 gives the patient household Euler Equation:
Cpt = Et
(Cp
t+1
βp(1 + iqt )
)(Hp
t+1
Hpt
)(1−η), (4.B.9)
and patient household housing demand:
1
Hpt Cjη
t
=
(η
1− η
)(Ph
t
Cpt Hp(1−η)
t
−βp(Ph
t+1 + rht+1)
Cpt+1Hp(1−η)
t+1
), (4.B.10)
and patient household money demand:
Cpt Hp(1−η)
t
(Mjt)
σm=
η
γm
(1− 1 + im
t
1 + iqt
). (4.B.11)
Impatient household Problem
Impatient household Lagrangian:
Li =E0
∞
∑t=0
(βi)t(
((Cit)
η(Hit)
1−η)1−σc
1− σc + γm (Mit)
1−σm
1− σm − γl (Lit)
1+σl
1 + σl
)+
λit
[(1 + i f
t−1)B ft−1 + (1 + ie
t−1)Bet−1 − (1 + im
t−1)Mit−1 − rh
t Hit−1 − wtLi
t−
rkt Kt − (1− θ)(Πtb
t + Πpt ) + Ci
t + It − B ft − Be
t + Mit + qh
t (Hit − Hi
t−1)
]−
λitq
kt
[Kt(1− δ) + It
(1− φ
2
(It
It−1− 1)2 )
− Kt+1
]+
µet
(B
e
t − qkt Kt
)+
µft
(B
f
t − qht Ht
). (4.B.12)
where λit is the Lagrangian multiplier associated with the impatient house-
hold budget constraint, qkt is the shadow value of physical capital associated
with the physical capital accumulation identity and µet and µ
ft denote the
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 167PDF page: 167PDF page: 167PDF page: 167
Mortgage loans and shadow banks 157
shadow value of the loan-to-value constraints. The FOCs w.r.t. Kt, It, B ft , Be
t ,
Cit, Hi
t, Mpt , Lp
t :
λitq
kt =λi
t+1rkt+1 + λi
t+1qkt+1(1− δ) + qk
t+1µet+1 (4.B.13)
λit =λi
tqkt
(1− φ
2
(It
It−1− 1)2
− φ
(It
It−1− 1)(
It
It−1
))+
λit+1qk
t+1βj
(φ
(It+1
It− 1)(
It+1
It
)2)
(4.B.14)
λit =λi
t+1βi(1 + i ft ) + µ
ft (4.B.15)
λit =λi
t+1βi(1 + iet) + µe
t (4.B.16)
λit =ηCi(η−1)
t (Ciηt Hi(1−η)
t )−σc(4.B.17)
λitq
ht =(1− η)Hi(−η)
t (Ciηt Hi(1−η)
t )−σc+ λi
t+1βi(qht+1 + rh
t+1) + µft qh
t
(4.B.18)
γm
(Mit)
σm =λit
(1− 1 + im
t1 + ie
t
)+ µe
t
(1 + im
t1 + ie
t
), (4.B.19)
γl(Lit)
σl=λi
t+1W it . (4.B.20)
Rewriting the FOCs, substituting out the Lagrangian multiplier λit and
µft and setting σc = 1 gives the patient household Euler Equation:
1
Ciηt Hi
t
=
(η
1− η
)[qh
t
CitH
i(1−η)t
−βi(qh
t+1 + rht+1)
Cit+1Hi(1−η)
t+1
+
qht
(βi(1 + i f
t )
Cit+1Hi(1−η)
t+1
− 1
CitH
i(1−η)t
)](4.B.21)
and money demand:
CitH
i(1−η)t
(Mit)
σm =η
γm
(1− 1 + im
t
1 + i ft
)+ µ
ft
(1 + im
t
1 + i ft
). (4.B.22)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 168PDF page: 168PDF page: 168PDF page: 168
158 Chapter 4
Firm Problem
Firm Lagrangian:
L f = rkt Kt−1 + wtLt − λ
ft
(Et(πs)At(Kt−1)
1−α(Lt)α −Yt
), (4.B.23)
where λft denotes firms’ marginal costs which are in a competitive environ-
ment equal to the price level. We obtain the following FOCs w.r.t. Kt−1 and
Lt:
rkt =
(1− α)Yt
Kt−1(4.B.24)
wt =αYt
Lt(4.B.25)
Assuming free-entry and exit, firms will enter until expected economic profits
are zero. Accordingly, firm profits:
Πt = Et(πs)(Yt −Wt(Lpt − Li
t)− rkt Kt), (4.B.26)
will be equal to zero.
4.C Bank optimization problem
The bank maximizes its profits subject to its budget constraint and equity
buffer constraint:
max(Etπs[(1 + i f
t )B f ,tbt + (1 + ie
t)Be,tbt ]− im,tb
t Mtbt − iq,tb
t Qtbt − B f ,tb
t −
Be,tbt − χ((1 + i f
t )B ft + (1 + ie
t)Bet )
−
λtbt (Be,tb
t + B f ,tbt + χ((1 + i f
t )B ft + (1 + ie
t)Bet )−Qtb
t −Mtbt )
µtbt ((1 + im,tb
t )Mtbt − πb[(1 + i f
t )B ft + (1 + ie
t)Bet ]), (4.C.27)
where λtbt and µtb
t are the Lagrangian multipliers associated with the budget
constraint and equity buffer constraint respectively. The FOCs w.r.t. B f ,tbt ,
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 169PDF page: 169PDF page: 169PDF page: 169
Mortgage loans and shadow banks 159
Be,tbt , Mtb
t , Qtbt are denoted by:
i ft =χ(1 + i f
t ) + λtbt (1 + χ(1 + i f
t )) + µtbt πb(1 + i f
t ), (4.C.28)
iet =χ(1 + i f
t ) + λtbt (1 + χ(1 + ie
t)) + µtbt πb(1 + ie
t), (4.C.29)
im,sbt =λtb
t + µtbt (1 + im,tb
t ), (4.C.30)
iq,sbt =λtb
t . (4.C.31)
From the FOCs we get that the shadow value with respect to the budget
constraint λtbt = iq,sb
t > 0. Consequently, we know that the budget constraint
holds with equality. Next we can combine the FOCs w.r.t. Mtbt and Qtb
t to
obtain: µtbt =
(im,sbt −iq,sb
t
1+im,tbt
). From the household problem we know that im,sb
t <
iq,sbt if γm > 0, i.e., when households value money, the bank equity buffer
constraints holds with equality. Combining the FOCs we obtain:
i ft − χ(1 + i f
t ) =λtbt (1 + χ(1 + i f
t )) + µtbt πb(1 + i f
t ), (4.C.32)
iet − χ(1 + i f
t ) =λtbt (1 + χ(1 + ie
t)) + µtbt πb(1 + ie
t). (4.C.33)
We can interpret these results as follows. Traditional banks will increase in-
vestment in either asset as long as the budget constraints and the equity
buffer constraints do not bind. It is possible to substitute the budget and
equity buffer constraints in the maximization problem to obtain:
max
Etπs[(1 + i ft )B f ,tb
t + (1 + iet)Be,tb
t ]− (4.C.34)
πb[(1 + i ft )B f
t + (1 + iet)Be
t ]− (1 + iq,tbt )Qtb
t
.
From this we know that as long as the return on a particular asset is larger
than the costs of deposit insurance and equity, traditional banks increase in-
vestment. Hence, the equity buffer constraint determines the amount of de-
posits, the credit supply curve is flat for lending rates larger than the costs of
deposit insurance and equity, and bank equity is determined as the residual
from the balance sheet identity.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 170PDF page: 170PDF page: 170PDF page: 170
160 Chapter 4
For shadow banks the problem is similar:
max(Etπs[(1 + i f
t )B f ,sbt + (1 + ie
t)Be,sbt ]− im,sb
t Msbt − iq,sb
t Qsbt − B f ,sb
t −
Be,sbt
− λsb
t (Be,sbt + B f ,sb
t −Qsbt −Msb
t )
µsbt ((1 + im,s
t )Mst − ν[k f
t (1 + i ft )B f
t + ket(1 + ie
t)Bet ]) (4.C.35)
where ν ≡ [P(d|H)πg + (1− P(d|H) − P(d|L))πb + P(d|L)πd]. The FOCs
w.r.t. B f ,sbt , Be,sb
t , Msbt , Qsb
t are denoted by:
Etπs(1 + i ft )− 1 = λsb
t + µsbt νk f
t (1 + i ft ) (4.C.36)
Etπs(1 + iet)− 1 = λsb
t + µsbt νke
t(1 + iet) (4.C.37)
im,sbt = λsb
t + µsbt (1 + im,sb
t ) (4.C.38)
iq,sbt = λsb
t . (4.C.39)
From the FOCs wrt Mtbt and Qtb
t we obtain again λsbt > 0 and µsb
t > 0 and so
both constraints hold with equality. Substituting out the multipliers gives:
Etπs(1 + i ft )− 1 =iq,sb
t +
(1 + i f
t
1 + im,sbt
)νk f
t (im,sbt − iq,sb
t ) (4.C.40)
Etπs(1 + iet)− 1 =iq,sb
t +
(1 + ie
t
1 + im,sbt
)νke
t(im,sbt − iq,sb
t ) (4.C.41)
and similar to the traditional banking problem we can substitute the con-
straints in the profit function to obtain:
max(Etπs[(1 + i f
t )B f ,sbt + (1 + ie
t)Be,sbt ]− ν[k f
t (1 + i ft )B f
t +
ket(1 + ie
t)Bet ]− (1 + iq,sb
t )Qsbt
(4.C.42)
From this we know that as long as the return on a particular asset is larger
than the expected liquidation costs and equity, traditional banks increase
investment. Hence, similar to traditional banks the equity buffer constraint
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 171PDF page: 171PDF page: 171PDF page: 171
Mortgage loans and shadow banks 161
determines the amount of deposits, the credit supply curve is flat for lending
rates larger than the costs liquidation insurance and equity, and bank equity
is determined as the residual from the balance sheet identity.
4.D Proof Proposition 4.1
For shadow banks the weighted average costs of funding is denoted by:
isbt =
qsbt
qsbt + msb
tiqt +
msbt
qsbt + msb
timt . (4.D.43)
Using the shadow bank balance sheet constraint (4.15) and the shadow bank
equity buffer constraint for asset ι (4.17):
isbt = iq
t +πb[κ
ιt(1 + iι
t)bιt]
bιt(1 + im
t )
(imt − iq
t)
. (4.D.44)
For traditional banks the weighted average costs of funding are:
itbt =
qtbt
qtbt + msb
tiqt +
mtbt
qt + mtbt
imt . (4.D.45)
Using the traditional bank balance sheet constraint (4.15) and the traditional
bank equity buffer constraint (4.16) :
itbt = iq
t +πb[(1 + iι
t)bιt]
(1 + imt ) (1 + χ) bι
t
(imt − iq
t)
. (4.D.46)
Equating the marginal costs of shadow banks (4.D.44) with the marginal
costs of traditional banks (4.D.46) to determine which banking sector has
higher marginal costs we obtain that both banking sectors have the same
marginal costs if:
11 + χ
= κιt. (4.D.47)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 172PDF page: 172PDF page: 172PDF page: 172
162 Chapter 4
From this we can conclude that both banks invest in both assets if:
11 + χ
= κet = κ
ft . (4.D.48)
Traditional banks invest only in corporate loans assets and shadow banks
invest only in mortgage loans if:
11 + χ
> κet and
11 + χ
< κft (4.D.49)
Traditional banks invest only in mortgage loans and shadow banks invest
only in corporate loans if:
11 + χ
< κet and
11 + χ
> κft . (4.D.50)
4.E Including idiosyncratic credit risk
In the main text we argued that both traditional and shadow banks always
completely diversify their portfolios if they have the opportunity to do so.
To diversify all idiosyncratic risk the bank must trade with other interme-
diaries as they are not able to completely diversify idiosyncratic risk by
themselves because it is, for example, costly (see Hanson et al. (2015)). To
diversify the risk of these projects, both types of banks trade in the interb-
ank market. Specifically, they sell Sι,it units of risky projects and they buy Bι,i
t
units of risky projects financed by other banks. Consequently, the actuarially
fair priced deposit guarantee system is expressed as:
Dt =
[P(d)P(d|L)(1− πd) + (P(d)(1− P(d|L)− P(u|L))+
P(u)P(u|H))(1− πb) + (P(u)P(u|H) + P(d)P(u|L)(1− πg)]πb+
P(d)P(d|L)πd(πb − πd)
]πb[(1 + iι
t)(I ιt − Sι
t))]+
P(d)P(d|L)[(1 + iιt)Bι
t](πb − πd), (4.E.51)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 173PDF page: 173PDF page: 173PDF page: 173
Mortgage loans and shadow banks 163
which states that the deposit insurance premium consists of two parts. The
first part calculates the probability of failure for the bank’s own investment
projects which are subject to both idiosyncratic and aggregate risk (I ιt − Sι
t),
see Figure 4.8 for the probabilities of these projects defaulting. Second, the
bank also owns (potentially) diversified securities Bιt. These securities are
not subject to idiosyncratic risk, but only to aggregate risk. The deposit in-
surance in this case only needs to cover the difference between the bad and
disaster state. Diversification lowers the premium paid for deposit insur-
ance and therefore allows traditional banks to invest more in the risky asset.
So, diversification allows banks to attract more deposits for a given amount
of equity. Therefore traditional banks will always completely diversify their
portfolio if diversification costs are sufficiently low.
Shadow banks do not gain directly from diversification. The shadow
bank equity buffer constraint is specified by:
[P(u|L)πg + (1− P(u|L)− P(d|L))πb+
P(d|L)πd][kιt(1 + iι
t)(I ιt + Bι
t − Sιt)] ≥ (1 + im,s
t )Mst , (4.E.52)
from which we learn that shadow banks do not gain directly from diversi-
fication as it does not impact the fundamental value of the asset. It is best to
read the participation constraint as a “worst case scenario outcome” which
is the occurrence of a pessimistic signal. Diversification does not matter as
it does not allow shadow banks to create additional risk-free debt claims.
However, shadow banks liquidate all their assets in case a signal about the
future state of the world is pessimistic. Traditional banks buy these assets,
but only if the price of the securities is fair:
κιt = (1− χ)
(Et|S′=L(πs)
)ϕι
1
(Bι
tBt
)ϕι
2
(Msb
tMt
), (4.E.53)
where χ is the DGS premium described by (4.E.52). It is evident from (4.E.52)
and (4.E.53) that χ is larger if the shadow banks sell non-diversified assets.
Consequently, the liquidation value is in expectation lower when shadow
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 174PDF page: 174PDF page: 174PDF page: 174
164 Chapter 4
banks have a non-diversified portfolio. For this reason, shadow banks also
diversify their portfolio completely when diversification costs are sufficiently
low.
If we assume a symmetric equilibrium (all banks are alike), S f ,it = B f ,i
t
and Se,it = Be,i
t so we can conclude that I f ,it = S f ,i
t = B f ,it and Ie,i
t = Se,it = Be,i
t
and we obtain the diversified optimization problem stated in the main text.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 175PDF page: 175PDF page: 175PDF page: 175
Chapter 5
Sectoral allocation and
macroeconomic imbalances in
EMU∗
5.1 Introduction
In the run-up to the introduction of the euro, both real and nominal interest
rates in the Southern members of the Economic and Monetary Union (EMU)
decreased markedly. This induced major capital flows from the North to
the South, which were initially considered to be largely benign.1 In retro-
spect however, the inflow of capital mainly fueled a boom of domestic lend-
ing and construction, contributing little to productivity growth or business
cycle convergence.2 As the discrepancy between the external debt level and
the capacity to repay kept growing, eventually the solvency of the recipient
regions came under pressure (see Giavazzi and Spaventa, 2010). Whereas
∗This chapter is based upon Gilbert and Pool (2016).1 See for instance Feldstein (2012) who describes the large intra-EMU capital flows and Blan-
chard and Giavazzi (2002) for a—at that time—common interpretation of these capital flows.2 Comunale and Hessel (2014) describe how the surge in domestic demand was the rooSt
cause behind the emergence of current account deficits. Fagan and Gaspar (2007) show thatcapital inflows fueled a consumption boom while Eichengreen (2010) and Holinski et al.(2012) show that the Southern countries became relatively less productive after monetaryintegration.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 176PDF page: 176PDF page: 176PDF page: 176
166 Chapter 5
there exists a fairly broad consensus regarding this narrative (see e.g. Bald-
win and Giavazzi, 2015), less is known about how the sectoral allocation of
capital came about. It is therefore also unclear whether the developments
in the first decade of EMU were an unfortunate one-off or something that
could have been foreseen and possibly prevented.
In this chapter, we first document empirically how the growth of the
nontradable sector in Southern Europe was a broad-based phenomenon ex-
tending beyond the construction- and real estate sectors. We then proceed by
constructing a tractable two-sector two-region general equilibrium model of
a monetary union. We simulate the non-linear transition path following the
permanent drop in the real interest rate experienced by Southern Europe in
the run-up to the introduction of the euro. The fall in the interest rate in-
duces a regional demand boom, which increases demand for both tradable
and nontradable goods. Whereas the nontradable sector is able to increase
prices and output, the tradable sector faces foreign competition and thus
has less room to increase prices. Therefore, in real terms, capital and labor
are cheaper in the nontradable sector and are (re)allocated to this sector.
In the North, Southern demand for tradables and upward pressure on the
EMU-wide interest rate induce wage moderation and a shift of resources to
the tradable sector. As such, cost competitiveness positions in the North and
the South diverge, while Southern external debt accumulates. Absent a debt-
elastic interest rate or a debt limit, there is nothing to stop this process. When
we extend the model to include a third region—the ‘Rest of the World’—the
effects of monetary integration in the Southern part of the union are ampli-
fied, while spillovers to the North are more muted, in part due to an appre-
ciation of the union’s exchange rate that limits the growth of the Northern
tradable sector.
We validate the model predictions empirically using a reduced form
Bayesian panel-VAR for 10 euro area countries. The key predictions hold
up well: countries which experienced negative interest rate shocks (relat-
ive to the euro area average), experienced faster growth of the nontradable
sector, but not faster growth of the tradable sector. As such, also empirically,
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 177PDF page: 177PDF page: 177PDF page: 177
Sectoral allocation and macroeconomic imbalances in EMU 167
negative interest rate shocks lead to growth of the relative size of the nontra-
dable sector. Negative interest rate shocks also contribute to a deteriorating
current account balance.
This chapter contributes to an emerging body of research that studies
the allocation of incoming capital flows in Southern Europe, both across
and within sectors, and the effects thereof on the external position and pro-
ductivity.3 Most closely related, Benigno and Fornaro (2014), Piton (2015)
and Kalantzis (2015), show that in a small open economy framework an
exogenous fall in the interest rate leads to (relative) growth of the nontra-
dable sector. Piton (2015) suggests that higher mark-ups in the nontradable
sector contribute to the relative growth of this sector, while Benigno and
Fornaro (2014) show how—in a setting where only the tradable sector ex-
periences productivity growth—the reallocation of labor to the nontradable
sector contributes to stagnating productivity growth. Kalantzis (2015) em-
phasizes how the interest rate drop results in both growth of the nontrad-
able sector as well as increasing leverage, which together make balance-of-
payments crises more likely.
The focus of this literature on small open economies models implies that
any feedback effects that might occur in a monetary union are omitted.4 We
show these feedback effects to be relevant, contributing to both wage mod-
eration and a shift of resources towards the tradable sector in North. In this
way, we also complement studies by Gadatsch et al. (2016) and Bettendorf
and Leon-Ledesma (2015), who focus on domestic drivers of the German
3 Reis (2013) focuses on financial frictions to show why relatively unproductive firms in thenontradable sector grow at the expense of the tradable sector. Gopinath et al. (2017) and Cec-chetti and Kharroubi (2015) show that financial frictions can contribute to the misallocationof capital within sectors, as capital is allocated to firms that have higher net worth but are notnecessarily more productive. Sy (2016) emphasizes how the interaction of a common mon-etary policy and heterogeneous inflation rates implies real rates that are lower in the Souththan in the North, contributing to growth of the Southern nontrable sector. To rationalize theboom-bust cycle experienced by much of the Eurozone, Ozhan (2017) shows how bank bal-ance sheets can amplify fluctuations that are driven by news on the valuation of non-tradedsector capital.
4 Over 1999-2007, the former high interest rate countries’ represented 32-36% of euro areaGDP and 40-41% of the euro area population, rendering the assumption that these countriescan be represented as small open economies within the euro area counterfactual. See alsoFagan and Gaspar (2007).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 178PDF page: 178PDF page: 178PDF page: 178
168 Chapter 5
current account surplus.
Moreover, while our model remains relatively tractable otherwise, we
allow for monopolistic competition and differences in productivity levels
between regions and sectors. This set-up allows us to show that the realloc-
ation towards the nontradable sector that follows the interest rate shock is
qualitatively invariant to the degree of competition in the nontradable sec-
tor and to differences in productivity levels across sectors. Thus, the realloc-
ation of resources also follows through when the nontradable sector is the
less competitive and/or less productive one. Accordingly, our results offer a
structural explanation for the empirical findings documented by Borio et al.
(2016) and Cette et al. (2016), that credit booms like those experienced by
Southern Europe after the introduction of the euro are associated with a
productivity slowdown driven by a reallocation of resources towards less
productive sectors.
The results in this chapter raise important policy issues, as to correct-
ing external imbalances and preventing new ones. The model suggests that
many of the developments in the first decade of EMU should not have come
as a surprise. Indeed, our model suggests that growth of the Southern non-
tradable sector, deteriorating competitiveness, and current account deficits
are all relatively straightforward consequences of the economic boom in-
duced by the sharp decline in real interest rates. A sufficiently strong reac-
tion of Southern interest rates to the accumulating debt, a leaning-against-
the-wind type of fiscal policy, or possibly macroprudential measures, could
have helped to moderate these developments, preventing the need for a
sharp rebalancing process later on.
In the absence of timely stabilizing measures, investors ‘waking up’ and
demanding a higher interest rate premium induces a sharp rebalancing pro-
cess during which Southern GDP falls. We investigate various policy op-
tions that can accommodate a less disruptive rebalancing process, focusing
on product market reforms that have the potential to both boost growth
and facilitate the rebalancing process. Firstly, we analyze the effects of a lib-
eralization of the Southern nontradable sector, i.e., allowing for more do-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 179PDF page: 179PDF page: 179PDF page: 179
Sectoral allocation and macroeconomic imbalances in EMU 169
mestic competition. Perhaps counter-intuitively, but in line with Cavelaars
(2006), this does not improve the region’s external position. As mark-ups
in the nontradable sector come down, demand for nontradable goods in-
creases and the sector expands. Total output in the South grows, while the
external position marginally deteriorates. Spillovers from liberalizing the
Northern nontradable sector are limited. Secondly, we simulate a decrease in
the mark-up on tradable goods (interpreted as a deepening of the European
internal market). This induces a shift of productive resources towards the
tradable sector and boosts growth, although in the short run it does come at
the expense of a deterioration of the external position of the union as whole.
The rest of the Chapter is organized as follows. Section 5.2 documents
stylized facts. Section 5.3 presents the model. Section 5.4 shows the simula-
tion results and Section 5.5 shows the empirical results. Section 5.6 discusses
policy option and Section 5.7 concludes.
5.2 Stylized facts
In anticipation of the introduction of the euro, nominal interest rates in
Southern Europe fell sharply. As this partly reflected falling inflation ex-
pectations, the drop of economically more relevant real interest rates was
less extreme. Nevertheless, the drop was substantial: in the three years prior
to the introduction of the euro, real one-year yields—the nominal one-year
yield on government debt minus Consensus inflation expectations one-year
hence—in Italy, Ireland, Portugal and Spain (the ‘IIPS’, with data for Greece
being unavailable before 1998) fell by on average four percentage points, see
Figure 5.1 panel a. Over the same period, real rates in the rest of the euro
area (REA) remained roughly constant.
In the first years of EMU, interest rates in the entire euro area increased.
Following the collapse of the dotcom bubble, interest rates came down again.
However, inflation expectations and realized inflation in the GIIPS remained
persistently above those in the REA. Consequently, real rates in the GIIPS re-
mained below those in the REA up to the onset of the crisis.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 180PDF page: 180PDF page: 180PDF page: 180
170 Chapter 5
Figure 5.1. Interest rates and macroeconomic imbalances
-300
-200
-100
0
100
200
300
400
1998 2000 2002 2004 2006
d: Current account (billion $)
100
120
140
160
180
200
1998 2000 2002 2004 2006
c: Export (1998 = 100)
-2
-1
0
1
2
3
4
5
1996 1998 2000 2001 2004 2006
a. Real interest rates (1y gov. bonds)
IIPS GIIPS REA
100
110
120
130
140
150
160
170
180
1998 2000 2002 2004 2006
b: Domestic demand (1998 = 100)
Notes: The IIPS include Ireland, Italy, Portugal and Spain, the GIIPS also includes Greece.The REA includes the other EMU-12 countries: Austria, Belgium, Finland, France, Germany,Luxembourg and the Netherlands. Figure 1 panel a shows the real 1 year interest rate, cal-culated as the 1 year yield on government bonds minus inflation expectations over the same1 year period (calculated using Consensus data). Figures 1 panels b and d are based on datafrom the IMF WEO database October 2015, figure 1c uses AMECO data.
Low and falling interest rates induced a domestic demand boom in the
GIIPS (Figure 5.1 panel b). Over 1999-2007, domestic demand in the GIIPS
grew by on average 3% a year. In the REA, domestic demand increased by
1.7% a year. The demand boom in the GIIPS contributed to a surge in im-
ports, but was not matched by a similar increase in exports. Export perform-
ance even somewhat lagged behind the REA (Figure 5.1 panel c). As a result,
the current account of the GIIPS which was balanced at the onset of EMU,
deteriorated sharply in the years thereafter. The GIIPS’ current account defi-
cit was matched by an increasing current account surplus in the REA (Figure
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 181PDF page: 181PDF page: 181PDF page: 181
Sectoral allocation and macroeconomic imbalances in EMU 171
5.1 panel d).5 Accordingly, the external position of the euro area as a whole
remained close to balance.
In Figure 5.2 we plot value added in the nontradable sector to show that
growth was mostly concentrated in the nontradable sector. To this end, we
use data from the World Input Ouput Database (Timmer et al., 2015) and
estimate for each sector in each country the share of production that is ab-
sorbed domestically. We aggregate these results for all euro area countries,
weighing each member by its share in total euro area output. Subsequently,
we construct the nontradable sector by selecting those sectors that depend
most heavily on domestic demand. Figure 5.A.1 in the Appendix shows for
the year 1999 per sector the share of production that is absorbed domestic-
ally. In Figure 5.2 panels a and b, we construct the nontradable sector by ag-
gregating the 8 sectors that depend most heavily on domestic demand and
which jointly produce 33% of total euro area output. In Figure 5.2 panels
c and d, we construct the nontradable sector by aggregating the 14 sectors
that depend most heavily on domestic demand and which jointly produce
50% of total output.
Figure 5.2 panel a and c show the growth of the nontradable and trade-
able sector. Irrespective of the threshold used, the nontradable sector in the
GIIPS realized higher growth rates than the nontradable sector in the REA.
This growth differential could be the result of higher GDP growth rates in
the GIIPS, i.e., higher growth in both the nontradable and tradable sectors,
potentially reflecting catch-up growth. However, Figure 5.2 panel b and d
show that this is not the dominant factor. Value added in the nontradable
sector as percentage of GDP grows in the GIIPS countries, while it decreases
in the REA. Hence, these latter countries realized predominantly export led
growth, while in the GIIPS the nontradable sector grew faster than the trad-
able sector.
Numerous country or sector specific reasons can be identified to explain
the allocation of capital inflows. One popular explanation focuses on exces-
5 Consistent with this pattern, Berger and Nitsch (2014) provide evidence of a significantwidening of bilateral intra-euro area trade imbalances.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 182PDF page: 182PDF page: 182PDF page: 182
172 Chapter 5
Figure 5.2. Nontradable sector growth and as percentage of GDP
a: Value added nontradable sector (33% of output) b: Value added nontradable sector as percentage of GDP
(33% of output)
c: Value added nontradable sector (50% of output) d: Value added nontradable sector as percentage of GDP
(50% of output)
e: Value added nontradable sector without construction f: Value added nontradable sector as percentage of GDP
(33% of output) without construction (33% of output)
Notes: GIIPS contain Greece, Ireland, Italy, Portugal, Spain. The REA contains the other EMU-12 countries excluding
Luxembourgh: Austria, Belgium, Finland, France, Germany and the Netherlands. In Figure a, c, e 1999 = 100.
80
100
120
140
160
180
200
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
21.0%
21.2%
21.4%
21.6%
21.8%
22.0%
22.2%
22.4%
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
29.0%
29.5%
30.0%
30.5%
31.0%
31.5%
1999 2000 2001 2002 2003 2004 2005 2006 2007 200880
100
120
140
160
180
200
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
80
100
120
140
160
180
200
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
14.5%
15.0%
15.5%
16.0%
16.5%
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
GIIPS REA
Notes: The GIIPS include Greece, Ireland, Italy, Portugal and Spain. The REA includes theother EMU-12 countries excluding Luxembourgh: Austria, Belgium, Finland, France, Ger-many and the Netherlands. In panels a, c and e 1999 = 100. Source: own calculations basedon WIOD, release 2013 (Timmer et al., 2015), see Appendix 5.A.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 183PDF page: 183PDF page: 183PDF page: 183
Sectoral allocation and macroeconomic imbalances in EMU 173
sive growth in the real estate sector. Housing bubbles have certainly been
an important factor driving current account imbalances in countries such
as Spain and Ireland. However, the nontradable boom was not limited to
real estate and construction. Figure 5.2 panels e and f show the growth rate
and share of GDP of value added in the nontradable sector excluding the
construction and real estate sector. Although growth rates are a bit lower
than in Figure 5.2 panel a, the overall pattern is the same. Thus, the rapid
growth of the nontradable sector appears to have been more broad-based
than is sometimes suggested.6
5.3 The model
The model builds on the two-region two-sector framework introduced by
Stockman and Tesar (1995) and Obstfeld and Rogoff (1995). The regions are
labeled ‘North’ and ‘South’. Following monetary integration, both regions
become part of a single monetary union. Both regions exist of a large number
of identical households, a large number of firms and a government which all
have perfect foresight. The union has a single central bank which keeps the
union price level constant. Households consume, supply labor, accumulate
financial assets (one-period risk free bonds), hold money, and own the firms.
Firms buy capital from capital producers and hire labor from households.
In each region there are two types of firms, producing nontradable goods
(N) and tradable goods (T) respectively. The tradable good is used either as
consumption good or as investment in the tradable and nontradable capital
stock. The nontradable good can only be consumed.
The monetary union as a whole is a closed economy, a simplifying as-
sumption which we relax in Section 5.4.2. Within the union labor is mo-
bile across sectors, but not between regions. Exchange rates are fixed, i.e.,
pegged in the immediate run-up to EMU, and irrevocably fixed thereafter.
In the run-up to EMU, regional interest rates are higher in South than in
6 The financial sector, another sector typically mentioned as a fast growing (closed) ‘services’sector, is too open to be part of our nontradable sector and thus not driving the growththereof.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 184PDF page: 184PDF page: 184PDF page: 184
174 Chapter 5
North by an exogenous premium, which can be thought of as reflecting
e.g. exchange rate or inflation risk (for a similar approach, see Kollmann
et al., 2015). Following the introduction of the euro and the establishment of
a single central bank, this premium disappears and interest rates converge.
5.3.1 Households
Households that live in region j ∈ n, s, where n = North and s = South,
maximize lifetime utility by choosing consumption, labor supply and money
holdings:
U j =∞
∑v=0
(βj)v
[log Cj
t −θ(Lj
t)1+σl
1 + σl
], (5.1)
θ, σl > 0 and 0 < βj < 1,
where Cjt denotes consumption in region j at time t and Lj
t denotes labor
supply. The parameters βj = 1/(1 + ρj), θ and σl denote the discount rate,
the weight of labor in the utility function and the inverse of the elasticity of
work effort, respectively.
The consumption good is a composite of a nontradable Cj,Nt and a trad-
able good Cj,Tt which are transformed into the final consumption good via
a standard aggregator function: Cjt =
(Cj,N
t
)η (Cj,T
t
)1−ηwhere 0 < η < 1
denotes the share of nontradables. Note that the tradable good is either pro-
duced in the home region j or in the foreign region denoted by j′, i.e. con-
sumption of the tradable good in region j is denoted as Cj,Tt = Cjj,T
t + Cjj′,Tt .
The nontradable good is only produced domestically. The consumer price
index is a composite of the price of the nontradable good Pj,Nt and the price
of the tradable good Pj,Tt and is obtained by minimizing the expenditure ne-
cessary to obtain one unit of the composite good Cjt , i.e., minimizing Pj
t Cjt =
∑jj′P
j′,Tt Cjj′,T
t + Cj,Nt Pj,N
t subject to the constraint Cjt = (Cj,N
t )η(Cj,Tt )1−η :
Pjt =
(Pj,N
t
)η (Pj,T
t
)1−η
(η)η (1− η)1−η. (5.2)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 185PDF page: 185PDF page: 185PDF page: 185
Sectoral allocation and macroeconomic imbalances in EMU 175
For the tradable good the law of one price holds, as there are no trade re-
strictions any price difference is arbitraged away: Pn,Tt = Ps,T
t .
Households can borrow or lend via single period bonds issued in both
North and South. We assume that, prior to EMU, there is an exogenous
wedge between Southern and Northern risk-free interest rates:
r f ,nt + ω = r f ,s
t , (5.3)
where r f ,jt is the endogenously determined risk free interest rate on bonds
issued by region j and ω is an exogenous premium that disappears after
monetary integration. The uncovered interest rate parity condition ensures
that after integration the nominal interest rate is the same in both regions:
r f ,nt = r f ,s
t ≡ r f ,et , where r f ,e
t is the union interest rate.7
It is a characteristic of international business cycle models with incom-
plete financial markets that there is no unique deterministic steady state, see
e.g. Schmitt-Grohe and Uribe (2003) and Boileau and Normandin (2008). In
particular, whereas the interest rate pins down both regions’ net lending,
their external asset holdings are indeterminate. To pin down the equilibria,
and prevent any one region from endlessly accumulating debt, we introduce
a debt-elastic interest rate premium xjt. The interest rate premium increases
in the regions’ external debt level:
xjt = ξe−N j
t − 1, (5.4)
where ξ denotes how strongly the interest rate premium responds to debt
accumulation and N jt ≡ NFAj
t/(Pj,Tt Y j,T
t + Pj,Nt Y j,N
t ) denotes the net foreign
asset position as percentage of GDP, NFAjt denotes the net financial assets
of region j and Pj,Tt Y j,T
t and Pj,Nt Y j,N
t denote nominal GDP in the tradable
and nontradable sector respectively. As such, a region’s borrowing rate is
given by rjt = r f ,j
t + xjt. This implies that the rate paid by the borrower is
7 As the union-wide price level is kept constant by the monetary authority, at the union levelthe nominal interest rate equals the real interest rate. This is not necessarily the case at thelevel of the individual regions however, as movements in relative prices can drive a wedgebetween nominal and real rates.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 186PDF page: 186PDF page: 186PDF page: 186
176 Chapter 5
higher than the one received by the lender. The difference can be thought of,
and microfounded as, the cost of financial intermediation (Boileau and Nor-
mandin, 2008). Alternatively, it can be interpreted as a premium on default
risk that is absorbed by the intermediary bearing the risk.8
The household budget constraints are represented by:9
j′
∑ Bj′ jt + Pj,T
t Cj,Tt + Pj,N
t Cj,Nt =
j′
∑(
1 + rjt−1
)Bj′ j
t−1 + πj,Nt + π
j,Tt + Lj
tWjt ,
(5.5)
where LjtW
jt denotes nominal labor income, Bj′ j
t denotes net bonds issued in
country j and held by households in country in j′, πj,Nt and π
j,Tt are firm
profits (hence households are the true owners of the firms). Households
maximize their utility by choosing consumption and labor supply and bond
holdings, subject to the budget constrained and a no-Ponzi condition. Labor
is perfectly mobile within regions, but does not move across the two re-
gions. As a consequence, the wage rate is equal across sectors but may differ
between regions.
5.3.2 Firms
In both regions the economy is occupied by two types of intermediate firms
which produce wholesale tradables (T) and wholesale nontradables (N), re-
spectively. For brevity we define Z ∈ (T, N). Intermediate firms hire labor
from the household sector, buy capital from the capital producers, and sell
their wholesale goods to retailers. Retailers use the wholesale goods to pro-
duce the final goods. The retailers are introduced only to realize monopol-
istic competition in a tractable manner.
8 During the first decade of EMU risk premia were mostly absent while they suddenlyspiked when the solvency of the Southern states became questionable. In Section 5.4.3 wediscuss the consequences of a sudden increase in the interest rate premium.
9 We assume that, within regions, actuarially fair priced state-contingent securities exist thatinsure each household against idiosyncratic variations in labor and dividend income. Con-sequently, at the regional level, individual household income will correspond to aggregatehousehold income.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 187PDF page: 187PDF page: 187PDF page: 187
Sectoral allocation and macroeconomic imbalances in EMU 177
The aggregate production technologies of the nontradable and tradable
intermediate firms are specified by a Cobb-Douglas form:
yj,Zt (i) = Aj,Z
t
(K j,Z
t−1(i))1−αZ (
Lj,Zt (i)
)αZ
, (5.6)
where Aj,Zt denotes the productivity level in region j and sector Z, K j,Z
t de-
notes the physical capital stock and αZ denotes the share of labor in produc-
tion. Total labor demand is given by Ljt = Lj,N
t + Lj,Tt . Both types of firms
accumulate capital according to the following accumulation identities:
K j,Zt+1 = (1− δ)K j,Z
t + I j,Zt , (5.7)
where I j,Zt denotes investment in the physical capital stock and δ is the de-
preciation rate. The nontradable and tradable intermediate firms minimize
their costs subject to their production constraint, see Appendix 5.B.
Capital producers sell their capital to the intermediate firms in a per-
fectly competitive environment. For reasons of tractability we assume that
capital producers acquire investment (mobile across borders and between
sectors) to produce capital. They borrow from the domestic households to
produce capital. Consequently, the return to capital equals the domestic
borrowing rate rjt. The nontradable and tradable capital production func-
tion is subject to diminishing returns to scale and represented by: I j,Zt −
φPj,Tt
2
(I j,Zt
K j,Zt− δ
)2
K j,Zt , where capital adjustment costs are denoted in the price
of tradables. Maximizing profits yields the price of capital, see Appendix
5.B.
We model monopolistic competition by introducing a retail sector that
aggregates the intermediate goods produced by the nontradable and trad-
able firms respectively, into two (tradable and nontradable) final goods. Re-
tailers buy the products of the intermediate firms and use the following CES
production functions to produce the final goods (Dixit and Stiglitz, 1977):
Y j,Zt =
[∫ 1
0yj,Z
t (i)1−1/µj,Zdi]1/(1−1/µj,Z)
, (5.8)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 188PDF page: 188PDF page: 188PDF page: 188
178 Chapter 5
where yj,Zt (i) denotes nontradable or tradable output produced by interme-
diate nontradable or tradable firm i, Y j,Zt is the final goods and µj,Z denotes
the degree of substitutability between the intermediate products and de-
termines the amount of market power of the nontradable and tradable firms.
In the limit (µj,Z → ∞), pricing is perfectly competitive. Hence, both the
nontradable and tradable sector intermediate firms face downward sloping
demand (5.8) for their products because we assume imperfect substitutabil-
ity between intermediate products.
Retailers minimize the cost of buying output from intermediate firms∫ 10 Pj,Z
t (i)Y j,Zt (i)di subject to the CES production function (5.8). The retail
sector is perfectly competitive. Therefore both type of retail firms maximize
their profit function by setting prices equal to their marginal costs mct(i).
The aggregate price of nontradable and tradable products can be expressed
as the weighted sum of the intermediate good prices:
Pj,Zt =
[∫ 1
0pj,Z
t (i)1−µj,Zdi]1/1−µj,Z
, (5.9)
where pj,Zt (i) is the price set by intermediate firm i for intermediate input
yj,Zt (i).
5.3.3 Monetary authority and government sector
As prices are flexible, monetary policy cannot affect the real allocation of re-
sources. However, nominal variables, in our case the nominal price level,
are affected by monetary policy and are indeterminate without a policy
rule. Given our monetary union set-up, we assume therefore that the monet-
ary authority stabilizes the union-wide price level Pet , which consists of the
weighted sum of the aggregate price levels of the two regions:
Pet = hPn
t + (1− h)Pst , (5.10)
where h denotes the respective share of North and (1− h) denotes the share
of South in the Union. As such, the aggregate price levels within the two re-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 189PDF page: 189PDF page: 189PDF page: 189
Sectoral allocation and macroeconomic imbalances in EMU 179
gions, as well as the price level of nontradable- and tradable products within
the regions, are allowed to fluctuate.
The government engages in debt-financed government consumption.
The government uses both tradable and nontradable goods to produce ag-
gregate government consumption Gjt . The aggregator function is similar to
the aggregator function for the private consumption good, and uses equal
weights:
Gjt =
(Gj,N
t
)η (Gj,T
t
)1−η, 0 < η < 1. (5.11)
The government minimizes the cost of a given amount of government con-
sumption: Pjt Gj
t = ∑jj′P
j′,Tt Gjj′,T
t + Gj,Nt Pj,N
t . It takes the prices of the trad-
able and nontradable goods as given. Accordingly, government consump-
tion of both goods depends on the respective prices levels of both goods.
While we set steady state government spending equal to zero, we do exper-
iment with debt-financed government spending (or saving) shocks.
5.3.4 Market equilibrium conditions
The goods market equilibrium in the market for nontradables requires that
production of nontradable goods in each region is equal to consumption of
nontradable goods of consumers and the government in each region:
Y j,Nt = Cj,N
t + Gj,Nt , (5.12)
where Gj,Nt denotes government spending in the nontradable sector. The
market for tradable goods and investment is fully internationally integrated.
Hence, equilibrium requires that in the Union as a whole production is equal
to consumption and investment:
∑j
Y j,Tt = ∑
j
[Cj,T
t + I j,Tt + I j,N
t + ACjt + ICj
t + Gj,Tt
]. (5.13)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 190PDF page: 190PDF page: 190PDF page: 190
180 Chapter 5
Here, ACjt = ∑Z
[φPj,T
t2
(I j,Zt
K j,Zt− δ
)2
K j,Zt
]denotes the combined capital ad-
justment costs in the tradable- and the nontradable sector (which, like the
investment good itself, is expressed in terms of tradables) and ICjt denotes
the cost of financial intermediation (xjtNFAj
t). While the current account of
the union as a whole thus needs to be balanced, individual regions are al-
lowed to run deficits or surpluses. As borrowing and lending is only pos-
sible through one-period risk free bonds, a region’s net financial asset posi-
tion (NFAt) is denoted by:
NFAjt =
(1 + r f ,j
t−1
)NFAj
t−1+
Pj,Tt
(Y j,T
t − Cj,Tt − I j,T
t − I j,Nt − ACj
t − ICjt − Gj,T
t
). (5.14)
The current account balance is defined as the first difference of a country’s
NFAt. Finally, equilibrium in the market for financial assets requires:
NFAnt + NFAs
t = 0. (5.15)
5.3.5 Calibration
We calibrate the model to match the evolution of the Northern and South-
ern parts of the euro area following monetary integration, and to simulate
the effect of various policy measures. Time is quarterly. The parameter val-
ues are presented in Table 5.1. Both regions are equal in size. For simplicity,
productivity levels are equal across regions and sectors, an assumption we
relax later on. We furthermore assume that the discount factor in the North
is higher than in the South. This assumption ensures that prior to monet-
ary integration, both the South and the North run close to balanced current
accounts. Following monetary integration and the resulting convergence of
interest rates, the South borrows from the North. We calibrate the size of
the country specific debt-elastic risk premium such that the South’s external
debt stabilizes at 70% of GDP, in line with the average external debt of the
GIIPS in 2007. There are two important differences between the tradable
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 191PDF page: 191PDF page: 191PDF page: 191
Sectoral allocation and macroeconomic imbalances in EMU 181
Table 5.1. Calibrated parameters
Parameters Description Value
βn Discount factor households, North 0.990βs Discount factor households, South 0.980σl Inverse of the elasticity of work effort 2.000θ Weight of leisure 1.000η Share of nontradables in consumption 0.667αT Share of labor in the tradable production function 0.550αN Share of labor in the nontradable production function 0.600δ Depreciation rate of physical capital 0.030µn,N Market power nontradable sector, North 5.000µs,N Market power nontradable sector, South 3.500µj,T Market power tradable sector, region j 10.000ξ Credit premium reaction 0.007Aj,Z Productivity region j, sector Z 1.000h Relative share of North in union 0.500φ Capital adjustment costs 2.000
and nontradable sector and between the nontradable firms in Northern and
Southern Europe which are introduced to mimic the stylized facts but have
no effect on the general results. First, we assume that αN > αT, i.e., the non-
tradable sector is more labor intensive than the tradable sector. Second, we
assume that µn,N > µs,N , i.e., the nontradable sector in the Southern part of
Europe is less competitive than the nontradable sector in the Northern part
of Europe. Third, in our baseline calibration the tradable sector is equally
competitive in both parts of the union (due to the existence of a single mar-
ket) and more competitive than the nontradable sector (that is, µj,T > µj,T∀j).
As we analyze a large and highly persistent (arguably permanent) shock
that can lead to large and long-lasting deviations from the initial steady
state, log-linearizing the model around the steady state can lead to mislead-
ing results. Instead, we carry out a numerical simulation of the full nonlinear
model, using Dynare’s deterministic setting (see Adjemian et al., 2011). This
assumes that i) the shock to interest rates is unexpected and ii) agents are
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 192PDF page: 192PDF page: 192PDF page: 192
182 Chapter 5
certain that no future shocks will occur (‘perfect foresight’). The key advant-
age of this approach is that it provides us with the exact transition path of
the endogenous variables following the shock to the Southern interest rate,
whereas any log-linearized solution becomes less accurate the further the
variables move away from their initial steady state.
5.4 Model simulations
5.4.1 Two-region model
After monetary integration, the interest premium paid by South disappears
and interest rates in North and South converge, see Figure 5.3 panel a. South
experiences a demand boom: households reduce saving and increase con-
sumption (Figure 5.3b), while firms increase investment. Capital starts to
flow from North to South.
The capital inflow needs to be allocated between the tradable and non-
tradable sector. The allocation depends on two main channels. First, as wages
increase (Figure 5.3c), the more labor intensive nontradable sector experi-
ences a relative cost increase compared to the tradable sector. The relative
price of the nontradable good thus increases. As a result, demand for non-
tradable products increases less than the demand for tradable products (‘de-
mand effect’). Southern consumption of tradables increases by more than
the consumption of nontradables.
However, whereas tradable goods can be imported, nontradable goods
need to be produced at home. Consequently, a second channel emerges
which more than offsets the first one. In the absence of foreign competition,
firms in the nontradable sector have relatively ample space to increase prices
when their production costs increase without slackening demand (‘supply
effect’). Competition with the North implies that Southern firms active in the
tradable sector have less room to increase their prices as production costs in-
crease. The rising relative price of nontradables implies that, in real terms,
capital and labor are cheaper inputs in the nontradable sector. The effect is
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 193PDF page: 193PDF page: 193PDF page: 193
Sectoral allocation and macroeconomic imbalances in EMU 183
Figure 5.3. Consequences of monetary integration
1.0%
1.5%
2.0%
2.5%
1 3 5 7 9 11 13 15
a: risk-free quarterly interest rate
RF (S) RF (N)
2.84
2.86
2.88
2.90
2.92
1.56
1.58
1.60
1.62
1.64
1 3 5 7 9 11 13 15
b: consumption
C (S) C (N r-axis)
1.32
1.33
1.34
1.35
1.36
1.12
1.13
1.14
1.15
1.16
1 3 5 7 9 11 13 15
c: wage developments
Wage (S) Wage (N r-axis)
0.84
0.86
0.88
0.90
0.92
0.94
0.73
0.74
0.75
0.76
1 3 5 7 9 11 13 15
d: sectoral allocation (North)
KN/KT (N) LN/LT (N r-axis)
0.98
1.00
1.02
1.04
1.06
1.08
0.79
0.80
0.81
0.82
1 3 5 7 9 11 13 15
e: sectoral allocation (South)
KN/KT (S) LN/LT (S r-axis)
0.65
0.66
0.67
0.68
0.69
0.70
0.74
0.75
0.76
0.77
0.78
0.79
1 3 5 7 9 11 13 15
f: relative sectoral size
YN/YT (S) YN/YT (N r-axis)
1.06
1.07
1.08
1.09
1.10
1.11
1 3 5 7 9 11 13 15
g: real exchange rate
Real exchange rate (P_S / P_N)
-10%
-8%
-6%
-4%
-2%
0%
-60%
-50%
-40%
-30%
-20%
-10%
0%
1 3 5 7 9 11 13 15
h: external position (% GDP)
NFA (S) CA (S r-axis)
1.0%
1.5%
2.0%
2.5%
1 3 5 7 9 11 13 15
i: risk-adjusted quarterly interest
rate
R (S) R (N)
Notes: The Figure shows the effects of the permanent elimination of the wedge ω betweenSouthern and Northern risk-free interest rates in Equation (5.3). The x-axis displays the num-ber of quarters following the shock.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 194PDF page: 194PDF page: 194PDF page: 194
184 Chapter 5
a reallocation of capital and labor towards the nontradable sector (Figure
5.3f).
In the North the opposite effect occurs: Southern demand for capital in-
creases interest rates, which causes households to increase savings. Whereas
Southern demand for tradables grows, domestic demand in the North falls.
As a result, the nontradable sector shrinks and both wages and the relative
price of nontradables fall. Capital and labor are reallocated to the growing
tradable sector.
The Southern boom in consumption and investment and the shift of pro-
ductive resources to the nontradable sector cause the external position of
South to deteriorate (Figure 5.3h). The increase in external debt causes an
increase in the risk premium until the interest rate reaches a level at which
the capital inflow stops and the net foreign asset position stabilizes. The
rising interest rate also facilitates a shift of resources back to the tradable
sector to produce the goods necessary to balance imports and exports.
All results are obtained under the assumption of equal productivity levels
across sectors and countries. This simplifies the interpretation of the results
(e.g. ensuring that a sectoral reallocation of resources does not itself affect
GDP), but is clearly not a realistic assumption. We therefore calibrate the
productivity levels in the tradable and nontradable sector in both regions
using the database constructed by Mano and Castillo (2015). Productivity
is calculated as total value added per sector and country divided by total
hours worked in each sector and country. Mano and Castillo (2015) classify
a sector as tradable if more than 10% of the sector is exported. We aggreg-
ate productivity at the region level by taking the weighted average based
on the countries share in total EMU valued added. The resulting productiv-
ity levels, where we normalize tradable productivity in the North to 1 are:
AN,n = 0.76, AT,n = 1, AN,s = 0.79, AT,s = 0.92.
Qualitatively, the results are unchanged: the reallocation to the nontra-
dable sector following monetary integration still follows through when the
nontradable sector is the less productive sector (results not included but
available on request). As such, even if productivity in both sectors would
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 195PDF page: 195PDF page: 195PDF page: 195
Sectoral allocation and macroeconomic imbalances in EMU 185
Figure 5.4. Consequences of monetary integration: effects of including theRoW
1.0%
1.5%
2.0%
2.5%
1 3 5 7 9 11 13 15
a: risk-free quarterly interest
rate (South)
RF RF
-80%
-60%
-40%
-20%
0%
-14%
-10%
-6%
-2%
2%
1 3 5 7 9 11 13 15
b: external position (South, %
GDP)
CA CA
NFA r-axis NFA r-axis
0.94
0.95
0.96
0.97
1 3 5 7 9 11 13 15
c: exchange rate RoW / EA
Exchange rate RoW/EA
1.50
1.52
1.54
1.56
1.58
1.60
1 3 5 7 9 11 13 15
d: relative prices (South)
PN/PT PN/PT
0.95
1.00
1.05
1.10
0.79
0.80
0.81
0.82
1 3 5 7 9 11 13 15
d: sectoral allocation (South)
KN/KT KN/KT
LN/LT r-axis LN/LT r-axis
0.80
0.85
0.90
0.95
0.73
0.74
0.75
0.76
1 3 5 7 9 11 13 15
f: sectoral allocation (North)
KN/KT KN/KT
LN/LT r-axis LN/LT r-axis
Notes: The Figure shows the effects of the permanent elimination of the wedge ω betweenSouthern and Northern risk-free interest rates in Equation (5.3) in a closed (2-region)- andopen (3-region) version of the model.
remain constant, the relative growth of the nontradable sector hurts aggreg-
ate productivity. Accordingly, the model offers a structural explanation for
the empirical findings documented by Borio et al. (2016) who show that
credit booms like those experienced by Southern Europe after the introduc-
tion of the EMU are associated with a productivity slowdown driven by a
reallocation of resources towards less productive sectors.
Results do also not depend qualitatively on the degree of competition in
the Southern nontradable sector (see Figure 5.D.2 in Appendix 5.D). Even
with a perfectly competitive nontradable sector, the fall in Southern interest
rates induces a reallocation towards the nontradable sector. As such, elimin-
ating ‘rent seeking’ does—in our setting at least—not prevent the allocation
of incoming capital flows towards the production of nontradables.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 196PDF page: 196PDF page: 196PDF page: 196
186 Chapter 5
5.4.2 Including the Rest of the World
So far, we assumed a closed economy for the monetary union as a whole.
In this section, this simplifying assumption is relaxed by including a third
country labeled ‘Rest of the World’ (RoW). The size of the labor force of
RoW is equal to the combined Northern and Southern part of the monet-
ary union. RoW has a flexible exchange rate with the monetary union, is
connected to (initially, the Northern part of) the monetary union via an un-
covered interest rate parity condition and the Law of One Price and in terms
of parameters mimics the Northern part of the monetary union. It is, there-
fore, best thought of as another advanced economy. See Appendix 5.C for
the technical details.
As before, we simulate an interest rate shock in South, see Figure 5.4.
The addition of a third region somewhat amplifies the effects of this shock
in South, while it attenuates the effects in North. Two channels are at work.
First, there is an exchange rate effect. The Southern boom increases the union-
wide risk-free rate and induces an appreciation of the union’s currency. To
remain competitive, prices of tradeable products must fall. In South, this
amplifies the relative price increase of the nontradable good which contrib-
utes to an even faster reallocation of resources towards the nontradable sec-
tor. In North, this mitigates the fall in the relative price of the nontradable
good, which dampens the reallocation towards the tradable sector. Secondly,
and somewhat trivially, the addition of a third region increases the size of
the total economy. Southern imports no longer need to come exclusively
from North. As a result, the impact of the Southern boom on interest rates
in North is attenuated. Interest rates do rise, and North continues to realize
a current account surplus, but this is only approximately half as large as in
the two-region case. In contrast, the attenuated response of risk-free interest
rates implies South enjoys a boom and a current account deficit which are
even larger than in the two-region case.
The Southern boom also induces the RoW to run a current account sur-
plus. Whereas the surplus in the Northern part of the union was induced
by a rising interest rate, the RoW surplus is induced by the appreciation of
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 197PDF page: 197PDF page: 197PDF page: 197
Sectoral allocation and macroeconomic imbalances in EMU 187
the union currency. RoW tradable goods get cheaper which means by the
Law of One Price that the domestic currency price increases. Consequently,
resources are reallocated to the tradable sector, the RoW enjoys a moderate
boom and realizes a surplus on its current account.
5.4.3 Crisis
Our model is primarily constructed, and calibrated, to examine how mon-
etary integration affects the external position and sectoral allocation of re-
sources in Southern Europe. Due to the presence of a debt-elastic interest
rate, the model is stable. Yet, even in this setting, it is fairly straightforward
to see how a crisis could occur. As in e.g. Eggertsson and Krugman (2012),
the crisis can be modeled as a ‘Minsky moment’ in which risk aversion sud-
denly increases. In our case, this is easiest to simulate through an unexpec-
ted, permanent increase in the elasticity of the risk premium to a region’s
debt level (see (5.4) in Section 5.3). The effects thereof are mostly intuitive
and the opposite of the ones presented in Section 5.1: external borrowing
and investment in the South collapse, consumption falls, and resources tem-
porarily reallocate to the tradable sector. Eventually a new steady state, with
a lower external debt level and a stable current account, is reached. Figure
5.D.3 in Appendix 5.D presents the results in more detail.
5.5 Empirical analysis
5.5.1 Methodology and data
In this chapter we motivate our model by the sharp decline in real interest
rate experienced by Southern Europe in the run-up to the introduction of
the euro. However, the predictions of our model are more general and can
be summarized as follows: a negative interest rate shock, e.g. as experienced
by multiple Southern European countries in anticipation of EMU, leads to a
reallocation of resources towards the nontradable sector and the emergence
of a current account deficit. The opposite holds for a positive interest rate
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 198PDF page: 198PDF page: 198PDF page: 198
188 Chapter 5
shock.
In order to verify this prediction empirically, we estimate a standard
macroeconomic framework with output growth, inflation and a short-term
interest rate which is often used to identify interest rate shocks (see, e.g.,
Svensson 1997 and Clarida et al. 1999). We, however, allow for three devi-
ations to fit the estimation more closely to our theoretical model. First, we
split output growth in nontradeable and tradeable output growth to exam-
ine separately the effect of an interest rate shock on the growth rates of each
sector. Second, as our model is formulated in real terms, we estimate the
model in real terms. Third, we add current account flows, thereby opening
up the model, as we are interested in the effect of interest rate shocks on
cross-border capital flows.
We estimate the following reduced form Bayesian panel-VAR system:
Xt =α0 + α1Dt + Φ(L)Xt + εt, (5.16)
where Φ(L) ≡ Φ1L1 + ... + ΦpLp is a lag polynomial and Xt is a vector
containing the observed variables as discussed:
Xt =
[(yN
t,i − yNt ), (y
Tt,i − yT
t ),(
Bt,i
Yt,i− Bt
Yt
), (ir
t,i − irt)
]′, (5.17)
where yNt,i denotes the growth rate of the nontradable sector at time t in coun-
try i from which we subtract the average growth rate of the nontradable
sector in the euro area, yNt , to control for any EA-wide trend, yT
t,i denotes the
growth rate of the tradable sector of which we subtract the average growth
rate of the tradable sector in the euro area, yTt , ir
t,i is the ex-ante expected real
interest rate of which we subtract the average real interest rate in the euro
area irt .
10 Finally, Bt,i/Yt,i denotes a country’s current account balance as per-
10 The use of time-fixed effects also controls for common trends at the euro area level, but inthat case the control group is a non-weighted average of developments in individual coun-tries. This places a disproportional weight on developments in small countries, and doesnot give an accurate picture of EA-wide developments. We experimented with time-fixedeffects by including annual dummy variables to our specification. Results are qualitativelythe same, see Figure 5.D.4 in Appendix 5.D.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 199PDF page: 199PDF page: 199PDF page: 199
Sectoral allocation and macroeconomic imbalances in EMU 189
centage of GDP of which we subtract the euro area average current account
balance Bt/Yt.11
If applicable we include an exogenous dummy variable denoted by the
vector of dummies Dt ≡[
D f ct , Dec
t
]′. The dummy D f c
t controls for the global
financial crisis taking the value 1 in 2008 and 2009 and zero otherwise and
the dummy Dect controls for the euro area crisis taking the value 1 in 2011
and 2012 and zero otherwise. Finally, εt is a vector of stacked reduced form
residuals.
To identify the shocks we assume that the structural shocks are ortho-
gonal and use a Cholesky decomposition. The ordering is as specified in
(5.17), i.e, we let the real interest rate adjust contemporaneously to nontra-
dable and tradable growth shocks, but growth in the nontradable and trad-
able sector is affected by a real interest rate shock only with a lag. The model
is estimated using a (pooled) Bayesian estimation procedure. The data is
observed at an annual frequency and therefore only one lag is included.
In order to let the data speak as much as possible, we impose a (agnostic)
Minnesota prior: all lag coefficients takes a prior value of 0.8. The hyper-
parameters are set at standard values, i.e., the overall tightness parameter is
set equal to 0.1 and the lag decay parameter is set equal to 1. As (5.16) also
includes two exogenous dummy variables we set the exogenous parameter
tightness to 100.
The growth rates for both the nontradeable and tradeable sector are cal-
culated using Eurostat data for countries for which disaggregated output
time series are available: Austria, Belgium, Germany, Finland, France, Italy,
Ireland, Netherlands, Spain and Portugal. The disaggregated output time
series are available on an annual basis and categorized in either the trad-
able or nontradable sector. Similar to the stylized facts presented in Figure
5.2 panel c and d, we construct the nontradable sector by aggregating the
14 sectors that depend most heavily on domestic demand and which jointly
11 All variables are expressed in growth rates or percentages and subsequently demeaned.These transformations ensure that our data series are stationary. Stationarity is also con-firmed by the Levin et al. (2002) panel unit root test.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 200PDF page: 200PDF page: 200PDF page: 200
190 Chapter 5
produce 50% of total output.12 Figure 5.A.1 in the Appendix shows which
sectors are classified as a nontradable sector. We use annual nominal in-
terest rates on 1-year government bonds as a proxy for the country-wide
nominal interest rate and the Consensus forecast inflation expectations one
year ahead to transform the nominal interest rates into ex-ante real interest
rates.13 Finally, we use data from the World Economic Outlook database to
collect data on current account balances.
The time series cover the period 1996-2013 as no disaggregated data is
available for all countries for the years 2014 and 2015. As in general no data
is available for Luxembourg, we drop this country from our sample. For
Greece we lack data on nominal interest rates on one-year governments
bonds before 1999. As our inflation expectations measure covers inflation
expectations over a one-year period, the only consistent way to create ex-
ante real interest rates is to use one-year interest rates. Table 5.D.3 in the
Appendix summarizes the descriptive statistics.14
Preferably we would like to estimate the full model using the identific-
ation restrictions as presented in Section 5.3, see, e.g., Smets and Wouters
(2003) and Christiano et al. (2005) for two seminal contributions to DSGE
estimation. However, given the limited availability of disaggregated output
data at the sectoral level, we resort to panel data and use a reduced form
estimation approach.
12 The Eurostat classification is slightly different from the WIOD classification presented inFigure 5.2. Specifically, the WIOD contains more detailed information about the opennessof the sectors, but data is only available until 2011. We therefore match the WIOD classi-fication with the Eurostat classification to categorize the Eurostat sectors in a tradable andnontradable sector, see Table 5.A.1 in the Appendix.13 We also estimated the model with both nominal interest rates and inflation expectations.Results, which are not presented here for conciseness, are qualitatively the same.14 For robustness we experiment with 10-year government bond yields as those are alsoavailable for Greece before 1999. The nominal rates are transformed in ex-ante expected realrates using the one-year inflation expectations. To do so we assume that inflation expect-ations remain constant over the 10-year period. Results, which are not presented here, arelargely consistent with the results presented below.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 201PDF page: 201PDF page: 201PDF page: 201
Sectoral allocation and macroeconomic imbalances in EMU 191
5.5.2 Empirical results
We estimate the Bayesian panel-VAR over the sub-period 1996-2008 and
over the entire sample period 1996-2013.15 The first regression covers the
build-up of euro area imbalances as explained by our model where the
second regression also includes the bust as simulated in Section 5.4.3. As
dynamics may differ between the build-up phase and the sudden bust, it is
informative to estimate the Bayesian panel-VAR over both the sub-period
and the entire period.
Figure 5.5 shows the impulse response functions following a positive
interest rate shock for the period 1996− 2008. In line with the model predic-
tions, a country that is hit by a positive interest rate shock of one standard
deviation experiences a decline in the growth rate of the nontradable sector.
Figure 5.5 shows that the same does not hold for the tradable sector. For one,
the impulse response function is smaller in magnitude, but more import-
antly, countries are less probable to experience any change in the growth
rate of the tradable sector as the credibility interval includes zero. However,
in contrast to the predictions of the model, a country’s current account is not
likely to be affected by an exogenous increase in the real interest rate.
As the model is symmetric we can also interpret negative interest rate
shocks, as experienced in Southern Europe. A back-of-the-envelope calcu-
lation shows that the 4 percent point (unconditional) decrease in Southern
European interest rates relative to Northern European interest rates (see Fig-
ure 5.1), caused, according to the impulse response functions, a relative in-
crease in nontradeable growth of about 1− 2% per annum for a period of
6− 8 years. Thereafter the impulse response function is slowly decreasing to
zero. This increase comes on top of the overall trend of increasing nontrade-
able sector growth observed in all European countries. Hence, it appears
that interest rate shocks can explain a large fraction of the higher growth
rates of the Southern European nontradeable sector described in Figure 5.2.
Figure 5.6 shows the results for the period 1996− 2013, which also in-
15 The Bayesian panel-VAR is estimated using the ECB BEAR-toolbox developed by Dieppeet al. (2016), which builds on the methodology surveyed by Canova and Ciccarelli (2013).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 202PDF page: 202PDF page: 202PDF page: 202
192 Chapter 5
Figure 5.5. Real interest rate shock for sample period 1996-2008
5 10 15 20
-0.3
-0.2
-0.1
0
5 10 15 20
-0.2
-0.1
0
5 10 15 20
0
0.2
0.4
Res
pons
e of
:
Shock:
5 10 15 20
-0.1
0
0.1
0.2
Notes: The black lines represent the median response to a real interest rate shock estimatedover the time period 1996-2008. Shaded areas denote 68% credibility intervals which are gen-erated by drawing 50, 000 draws from the posterior distribution of which 40, 000 draws arediscarded as burn-in iterations. Horizontal axes specify years. Vertical axes denote percentpoint deviations from average euro area growth, ratio or rate.
cludes the bust period. The response of nontradable and tradable sector
growth rates to an interest rate shock is now more markedly divergent.
A positive interest rate shock still causes the nontradable sector growth to
slow. However, the effect of the interest rate shock on tradable sector growth
is positive and at longer horizons marginally different from zero at the 68%
credibility interval. These results are in line with the model which predicts
a small increase in the growth rate of the tradeable sector following a posit-
ive interest rate shock. They also help explain part of the strong growth of
Northern Europe’s tradeable sector.
In the bust phase, a positive real interest rate shock also has a positive
impact on the current account balance. A rising interest rate relative to the
euro area average is associated with a sharply improving current account
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 203PDF page: 203PDF page: 203PDF page: 203
Sectoral allocation and macroeconomic imbalances in EMU 193
Figure 5.6. Real interest rate shock for sample period 1996-2013
5 10 15 20
-0.2
-0.1
0
5 10 15 20
-0.1
0
0.1
0.2
5 10 15 20
0
0.2
0.4
5 10 15 20-0.2
0
0.2
0.4
0.6
0.8
Res
pons
e of
:
Shock:
Notes: The black lines represent the median response to a real interest rate shock estimatedover the time period 1996-2013. Shaded areas denote 68% credibility intervals which are gen-erated by drawing 50, 000 draws from the posterior distribution of which 40, 000 draws arediscarded as burn-in iterations. Horizontal axes specify years. Vertical axes denote percentpoint deviations from average euro area growth, ratio or rate.
balance, in line with a ’sudden stop’ pattern.
5.6 Policy options and discussion
The results presented highlight major challenges in terms of correcting ex-
isting imbalances and preventing new ones. Macroprudential policy limit-
ing private sector borrowing could play an important role in preventing the
developments stressed in this chapter from reoccuring. Fiscal policy also
offers a fairly straightforward tool to lean against excessive private borrow-
ing. Fiscal consolidation curtailing domestic demand directly improves the
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 204PDF page: 204PDF page: 204PDF page: 204
194 Chapter 5
external position.16 The fall in domestic demand also reduces the relative
price of nontradables, inducing a shift of productive resources towards the
tradable sector.
Currently, however, the first challenge for EMU is to reduce existing im-
balances in a way that does not unduly harm GDP growth. In Section 5.4.3
we showed that a sudden increase in the interest rate premium induces
a sharp rebalancing process as we saw in practice. In this section various
policy options that can accommodate a less disruptive rebalancing process
are examined.
5.6.1 Increasing competition in the nontradable sector
Figure 5.7 shows the effects of a liberalization of the nontradable sector in
South, i.e., a decrease in the mark-up on nontradable products. A liberal-
ization of the nontradable sector causes prices of nontradeable products
to fall, increasing relative demand for nontradables. Real income also in-
creases, contributing to increased demand for both tradable and nontrad-
able products. As nontradable products need to be produced at home, this
leads to an expansion of the nontradable sector. The domestic shortage of
tradable products is imported from the North. Overall, output and the relat-
ive size of the nontradable sector increase while the current account position
deteriorates.
Spillovers from a liberalization of the Northern nontradable sector are
limited. The North grows and from a Southern perspective both external
demand and the interest rate increase. GDP and the sectoral allocation of re-
sources in the South are largely unaffected. The Northern reforms do induce
a fall in the Northern price level, which—given that prices at the union level
are held constant by the single central bank—temporarily allows for some
inflation in the South. This improves the ratio of net financial assets to GDP.
Figure 5.D.5 in Appendix 5.D displays the results in more detail.
16 In a monetary union, even away from the zero lower bound, Ricardian equivalence breaksdown due to the fact that there is only a limited reaction of the union-wide interest rateto fiscal policy in an individual country/region. As such, private borrowing does not fullyoffset government savings, and the external position improves.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 205PDF page: 205PDF page: 205PDF page: 205
Sectoral allocation and macroeconomic imbalances in EMU 195
Figure 5.7. Product market reform in South, transition path
1.82
1.83
1.84
1.85
1.86
1.55
1.56
1.57
1.58
1.59
1 3 5 7 9 11 13 15
b: private consumption
C (S) C (N r-axis)
1.37
1.42
1.47
1.52
1 3 5 7 9 11 13 15
a: relative price of nontradables
PN/PT (S) PN/ PT (N)
0.69
0.71
0.73
0.75
0.77
1 3 5 7 9 11 13 15
d: relative sectoral size
YN/YT (S) YN/YT (N)
5.04
5.06
5.08
5.10
5.12
4.06
4.08
4.10
4.12
4.14
1 3 5 7 9 11 13 15
c: GDP
GDP (S) GDP (N, r-axis)
-0.3%
-0.2%
-0.1%
0.0%
0.1%
-72%
-71%
-70%
1 3 5 7 9 11 13 15
f: external position (% GDP)
NFA (S) CA (S r-axis)
0.94
0.96
0.98
1
0.76
0.78
0.8
0.82
1 3 5 7 9 11 13 15
e: sectoral allocation
KN/KT (S) LN/LT (S r-axis)
Notes: This Figure shows the effects of a permanent 10 percentage points reduction of mark-ups in the Southern nontradable sector. Simulation conducted using the 2-region version ofthe model; results using the 3-region version are highly similar and available upon request.
5.6.2 Deepening the internal market
The introduction of the euro was intended in part to deepen the internal
market, thereby increasing competition in the market for tradables. Evid-
ence on whether the euro achieved this is mixed. Deepening the internal
market is however still seen as a policy priority, see e.g. European Com-
mission (2015). We simulate the effects of a deepening of the internal market
through a decrease in the mark-up on tradables in both regions of the EA. As
shown in Figure 5.8, this induces a fall in the relative price of tradables and
thereby speeds up the desired shift of resources towards the tradable sector.
It boosts investment and GDP growth. As demand for tradable goods in-
creases faster than supply, the EA initially develops a trade deficit with the
RoW. This is accommodated by an appreciation of the euro, which allows
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 206PDF page: 206PDF page: 206PDF page: 206
196 Chapter 5
Figure 5.8. Deepening the EA internal market, transition path
1.32
1.34
1.36
1.38
1.40
1.42
1.44
1.46
1.50
1.55
1.60
1.65
1 3 5 7 9 11 13 15
a: relative price of nontradables
PN/PT (S) PN/ PT (S r-axis)
0.60
0.62
0.64
0.66
0.68
0.70
0.72
1 3 5 7 9 11 13 15
d: relative sectoral size
YN/YT (S) YN/ YT (N)
1.80
1.81
1.82
1.83
1.84
1.53
1.54
1.55
1.56
1.57
1 3 5 7 9 11 13 15
b: private consumption
C (S) C (N r-axis)
-2.0%
-1.5%
-1.0%
-0.5%
0.0%
0.5%
-90%
-86%
-82%
-78%
1 3 5 7 9 11 13 15
e: external position South (%
GDP)
NFA (S) CA (S r-axis)
5.10
5.15
5.20
5.25
5.30
4.04
4.08
4.12
4.16
4.20
4.24
1 3 5 7 9 11 13 15
c: GDP
GDP (S) GDP (N, r-axis)
0.92
0.94
0.96
0.98
1 3 5 7 9 11 13 15 17 19
f: exchange rate RoW / EA
Exchange rate RoW/EA
Notes: This Figure shows the effects of a permanent reduction of mark-ups in the Northernand Southern tradable sector.
for a rise in the price of tradable goods in the RoW that dampens local de-
mand. Over the long run, the tradable sector in the RoW shrinks marginally,
whereas the tradable sector in both regions of the EA grows significantly.
5.7 Conclusion
In this chapter, we documented empirically how growth of the nontradable
sector in Southern Europe was a broad-based phenomenon that extended
beyond the construction- and real estate sectors. We then showed in a two-
region two-sector general equilibrium model that many of the key character-
istics of the first decade of EMU can be explained by the major interest rate
shock the Southern European countries experienced when joining EMU. The
interest rate shock can be shown to explain both the divergence of current
account positions and wage rates between Northern and Southern Europe,
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 207PDF page: 207PDF page: 207PDF page: 207
Sectoral allocation and macroeconomic imbalances in EMU 197
as well as the allocation of capital and labor towards the nontradable sector
in South. The allocation of incoming capital to the nontradable sector occurs
irrespective of any differences in competitiveness or productivity across sec-
tors or regions. We confirmed the relation between interest rate shocks and
growth of the nontradable sector in a Bayesian panel-VAR for 10 euro area
countries over the period 1996− 2013.
Our results highlight several challenges for policy makers. When for-
eign borrowing is not matched by an increased export capacity, a point
made forcefully by Giavazzi and Spaventa (2010), solvability problems can
emerge. In our model a debt-elastic interest rate prevents these solvabil-
ity problems from occurring. In reality, the reaction of interest rates to the
external debt level was arguably absent. Fiscal policy would have offered
a fairly straightforward tool to lean against excessive private borrowing.
When imposed after the collapse of the boom, increased public savings can
still help speed up the reallocation of productive resources towards the trad-
able sector, but only at the cost of an even deeper recession. In a quest for
a more desirable solution, we investigated two options for product market
reform. Improving the European internal market for tradables, i.e., further
strengthening competition in this sector, appears to be the most promising
option, facilitating a further rebalancing towards tradables while simultan-
eously boosting growth.
Our findings suggest various directions for further research. Through-
out this chapter, we have assumed that goods are either tradable or non-
tradable. From a policy perspective, it is a highly relevant question to what
extent improving the European internal market for services can contribute
to increasing the share of traded ‘goods’, and what effects this would have.
Additionally, our study focuses on the sectoral allocation of capital inflows
in a nearly frictionless environment. Others have focused on the allocation
of capital within sectors, highlighting the role of financial frictions. Combin-
ing both perspectives seems a fruitful avenue for further research. Finally,
our study points to the need of explicitly monitoring a country’s foreign bor-
rowing, which is done via the Macroeconomic Imbalances Procedure (MIP)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 208PDF page: 208PDF page: 208PDF page: 208
198 Chapter 5
since 2011. To enforce the MIP, instruments are needed to curtail excessive
borrowing. Macroprudential policy offers promise in this respect, but still
faces major challenges that need further investigation.
5.A Sectoral dependence on domestic demand
Figure 5.A.1. Share of value added from domestic demand in the euro area
0 0.2 0.4 0.6 0.8 1
Water Transport
Chemicals and Chemical Products
Basic Metals and Fabricated MetalRubber and Plastics
Air TransportElectrical and Optical Equipment
Mining and QuarryingPulp, Paper, Paper , Printing and Publishing
Wood and Products of Wood and Cork
Other Non-Metallic MineralOther Supporting and Auxiliary Transport Activities; Activities of…
Transport EquipmentCoke, Refined Petroleum and Nuclear Fuel
Textiles and Textile ProductsMachinery, Nec
Renting of M&Eq and Other Business Activities
Inland TransportFinancial Intermediation
Leather, Leather and FootwearWholesale Trade and Commission Trade, Except of Motor Vehicles and…
Post and Telecommunications
Agriculture, Hunting, Forestry and FishingElectricity, Gas and Water Supply
Manufacturing, Nec; RecyclingRetail Trade, Except of Motor Vehicles and Motorcycles; Repair of…
Sale, Maintenance and Repair of Motor Vehicles and Motorcycles;…Other Community, Social and Personal Services
Food, Beverages and Tobacco
Hotels and RestaurantsReal Estate Activities
ConstructionEducation
Public Admin and Defence; Compulsory Social SecurityHealth and Social Work
Private Households with Employed Persons
Notes: The red-bar sectors sum to a nontradable sector that produces 33% of total euro areaoutput and the red- and yellow-bar sectors sum to a nontradable sector that produces 50%of total euro area output. Source: own calculations using WIOD, release 2013 (Timmer et al.,2015).
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 209PDF page: 209PDF page: 209PDF page: 209
Sectoral allocation and macroeconomic imbalances in EMU 199
Table 5.A.1. Classification
WIOD Classification VA Dom- Total Cumulative EUROSTATestically output share & KLEMS
codes
Private Households with Employed Persons 100.0% 26194 0.2% THealth and Social Work 99.8% 574381 4.7% QEducation 99.7% 378891 7.7% PReal Estate Activities 99.7% 890490 14.6% LPublic Admin and Defence; 99.5% 603584 19.4% O
Compulsory Social SecuritySale, Maintenance and Repair of Motor 99.0% 223209 21.1% G45 - G46 -
Vehicles and Motorcycles - G47Construction 98.6% 913288 28.3% FHotels and Restaurants 98.5% 359423 31.1% IRetail Trade, Except of Motor Vehicles and 98.1% 485831 34.9% G47
Motorcycles; Repair of Household Goods
Other Community, Social and Personal Serv. 97.2% 429865 38.2% R + S + UElectricity, Gas and Water Supply 97.0% 271372 40.4% DPost and Telecommunications 94.2% 255341 42.4% J - J62-63
+ H53Wholesale Trade and Commission Trade, 92.6% 643040 47.4% G46
except of motor vehicles and motorcyclesFinancial Intermediation 92.4% 614538 52.2% K
Inland Transport 90.6% 292791 54.5% H - H53Renting of M&Eq and Other 90.4% 1144176 63.5% M + N
Business Activities + J 62-63Other Supporting and Auxiliary Transport 85.5% 238207 65.3% H - H53
Activities; Activities of Travel AgenciesAgriculture, Hunting, Forestry and Fishing 84.0% 315989 67.8% AMining and Quarrying 83.2% 76820 68.4% BWood and Products of Wood and Cork 79.8% 80283 69.0% CFood, Beverages and Tobacco 77.1% 568989 73.5% COther Non-Metallic Mineral 76.5% 151818 74.7% CPulp, Paper, Printing and Publishing 74.4% 285858 76.9% CCoke, Refined Petroleum and Nuclear Fuel 73.6% 127712 77.9% CBasic Metals and Fabricated Metal 68.7% 475078 81.6% CManufacturing, Nec; Recycling 66.3% 138993 82.7% CRubber and Plastics 62.9% 152780 83.9% CAir Transport 60.5% 74950 84.5% H - H53Textiles and Textile Products 54.3% 183217 85.9% CLeather, Leather and Footwear 52.7% 46897 86.3% CMachinery, Nec 46.9% 366348 89.2% CElectrical and Optical Equipment 44.5% 416984 92.4% CTransport Equipment 42.7% 533056 96.6% CChemicals and Chemical Products 40.8% 391655 99.7% CWater Transport 26.7% 40050 100.0% H - H53
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 210PDF page: 210PDF page: 210PDF page: 210
200 Chapter 5
5.B Model solution
Households
Household maximization problem:
max∞
∑v=0
(βj)v
log (Cj,Nt )η(Cj,T
t )1−η − θ(Ljt)
1+σl
1 + σl+
χ(
mjt
)1−σm
1− σm
s.t.
j′
∑j
Bjt + Pj,T
t Cj,Tt + Pj,N
t Cj,Nt + Mj
t =j′
∑j(1 + rj
t−1)Bjt−1+
πj,Nt + π
j,Tt + Lj
tWjt + Mj
t−1. (5.B.1)
Households maximize their utility by choosing both consumption goods,
labor supply, money holding and bond holdings, subject to the budget con-
straint and a no-Ponzi condition. The FOCs are:
1− η
Cj,Tt
= Pj,Tt λh
t , (5.B.2)
η
Cj,Nt
= Pj,Nt λh
t , (5.B.3)
λht = λh
t+1(1 + rjt)βj, (5.B.4)
θ(Ljt)
σl = W jt λh
t , (5.B.5)
χ(
mjt
)−σm= λh
t − λht+1βj, (5.B.6)
where λht denotes the households’ Lagrangian multiplier. Using the FOC
for the tradable consumption good (5.B.2), Cj,Tt , to substitute the Lagrangian
multiplier out gives:
Pj,Tt Cj,T
t =Pj,T
t+1Cj,Tt+1
(1 + rjt)βj
, (5.B.7)
1− η
Cj,Tt Pj,T
t
=η
Cj,Nt Pj,N
t
, (5.B.8)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 211PDF page: 211PDF page: 211PDF page: 211
Sectoral allocation and macroeconomic imbalances in EMU 201
Cj,Tt Pj,T
t =1− η
θ
W jt
(Ljt)
σl, (5.B.9)
Mjt
Pjt
=
[χ
Cj,Tt Pj,T
t1− η
(1 + rj
t
rjt
)] 1σm
. (5.B.10)
Firms
Retailers are perfectly competitive and therefore we consider a representat-
ive retailer which buys input yj,Zt (i) from intermediate firm i and produces
output Y j,Zt according the following aggregator function:
Y j,Zt =
[∫ 1
0yj,Z
t (i)(µj,Z−1)/µj,Z
di]µj,Z/(µj,Z−1)
, (5.B.11)
and has a budget constraint which is denoted by: Pj,Zt Y j,Z
t =∫ 1
0 pj,Zt (i)yj,Z
t (i)di.
Retailers minimize their cost subject to their production function:
Ljt =
∫ 1
0pj,Z
t (i)yj,Zt (i)di−
λrt
(Y j,Z
t −[∫ 1
0yj,Z
t (i)(µj,Z−1)/µj,Z
di]µj,Z/(µj,Z−1)
), (5.B.12)
where λrt is the retailers marginal cost of producing an extra unit of final
output. The FOC w.r.t. to production input yj,Zt (i) of firm i and production
input yj,Zt (i′) of firm i′ and dividing these FOCs gives the relative pricing
equation:
yj,Zt (i) = yj,Z
t (i′)
(pj,Z
t (i)
pj,Zt (i′)
)−µj,Z
. (5.B.13)
If we combine the budget identity Pj,Zt Y j,Z
t =∫ 1
0 pj,Zt (i)yj,Z
t (i)di and aggreg-
ator function (5.B.11) and substitute subsequently the relative pricing equa-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 212PDF page: 212PDF page: 212PDF page: 212
202 Chapter 5
tion (5.B.13) to solve for Pj,Zt , we obtain:
Pj,Zt =
[∫ 1
0pj,Z
t (i)1−µj,Zdi]1/1−µj,Z
. (5.B.14)
We can substitute (5.B.14), together with the relative pricing equation (5.B.13)
for yj,Zt (i), back in the budget identity to obtain retailer demand for interme-
diate product yj,Zt (i):
yj,Zt (i) =
[Pj,Z
t
pj,Zt (i)
]µj,Z
Y j,Zt . (5.B.15)
Intermediary firms minimize their costs which consists of unit labor costs,
the opportunity costs of holding a unit of capital and the costs of buying
capital from the capital producers. The intermediate firm Lagrangian is ex-
pressed by:
Lj,Zt (i) =W j
t Lj,Zt (i) + (1 + rj
t)qj,Zt K j,Z
t (i)− qj,Zt (1− δj)K j,N
t (i)
− λj,Zt (i)
(Aj,Z
t (K j,Zt (i))1−αj
(Lj,Zt (i))αj
), (5.B.16)
where λj,Zt (i) is the Lagrangian multiplier of the firms which represents the
intermediate firms’ marginal costs. The FOC w.r.t. Lj,Zt (i) is represented by:
W jt = λ
j,Zt (i)
αjyj,Zt (i)
Lj,Zt (i)
, (5.B.17)
where yj,Zt (i) = Aj,Z
t (K j,Zt (i))1−αj
(Lj,Zt (i))αj
. The FOC w.r.t. K j,Zt (i) is repres-
ented by:
1 + rjt =
λj,Zt (i) (1−αj)yj,Z
t (i)K j,Z
t (i)+ (1− δ)qj,Z
t−1
qj,Zt
. (5.B.18)
Using both FOCs in the production function gives the expression for mar-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 213PDF page: 213PDF page: 213PDF page: 213
Sectoral allocation and macroeconomic imbalances in EMU 203
ginal costs, λj,Zt , which is the same for all intermediate firms:
λj,Zt =
1
Aj,Zt
((1 + rj
t)qj,Zt + (1− δ)qj,Z
t−1
(1− αj)
)(1−αj)(W j
tαj
)αj
. (5.B.19)
Capital producers are perfectly competitive. Consequently, the individual
capital producer’s optimization problem corresponds to the aggregate prob-
lem. The aggregate capital stock evolves according to:
K j,Zt+1 = (1− δj)K j,Z
t + I j,Zt . (5.B.20)
Capital producers combine new investment I j,Zt with undepreciated cap-
ital to produce new capital according to the following production function:
Φ(
I j,Zt
K j,Zt
)K j,Z
t = I j,Zt −
φPj,Tt
2
(I j,Zt
K j,Zt− δ
)2
K j,Zt . This functional form ensures
that the price of capital is equal to unity in steady state. Capital producers
choose investment I j,Zt to maximize the profits from producing capital and
then sell their capital for a price qj,Zt :
maxI j,Zt
qj,Z
t
[I j,Zt −
φPj,Tt
2
(I j,Zt
K j,Zt
− δ
)2
K j,Zt
]− Pj,T
t I j,Zt
. (5.B.21)
The FOC gives the price of capital:
qj,Zt = Pj,T
t
[1 + φ
(I j,Zt
K j,Zt
− δ
)]. (5.B.22)
As retailers face imperfect substitutability between intermediate inputs, in-
termediate firms have some market power and can set their prices as a mark
up over their marginal costs λj,Zt . Intermediary firms maximize their profits
w.r.t. prices:
πj,Zt (i) = [pj,Z
t (i)− λj,Zt (i)]yj,Z
t (i), (5.B.23)
The FOC w.r.t. pj,Zt (i) after we have substituted demand for yj,Z
t (i) (5.B.15)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 214PDF page: 214PDF page: 214PDF page: 214
204 Chapter 5
and solving for pj,Zt (i) gives:
pj,Zt (i) =
(µj,Z
µj,Z − 1
)λ
j,Zt . (5.B.24)
Hence, intermediate firms set prices as a mark-up over their marginal costs.
We can subsequently use (5.B.11) and (5.B.14) to aggregate over all firms
and rewrite (5.B.17), (5.B.18) and (5.6) in terms of aggregate output Y j,Zt and
aggregate prices Pj,Zt .
Government sector
The aggregator function for the public consumption good is similar to the
aggregator function for the private consumption good, with equal weights:
Gjt = (Gj,N
t )η(Gj,Tt )1−η , 0 < η < 1. (5.B.25)
The government minimizes the cost of a given amount of government con-
sumption: Pjt Gj
t = ∑jj′P
j′,Tt Gjj′,T
t + Gj,Nt Pj,N
t . It takes the prices of the trad-
able and nontradable goods as given. Accordingly, government consump-
tion of both goods depends on the respective prices levels of both goods:
Gj,Tt
Gj,Nt
=
(1− η
η
)Pj,N
t
Pj,Tt
. (5.B.26)
5.C Including the Rest of the World
The RoW economy is set up the same way as the Northern and Southern re-
gion, but has its own (floating) exchange rate. As before, the various regions
are denoted by superscript j, with j ∈ n, s, r.Prior to monetary integration, we assume the Rest of the World to be
connected to the Northern part of the euro area via an UIP:
1 + rnf ,t = (1 + rr
f ,t)Er,n
t+1
Er,nt
, (5.C.27)
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 215PDF page: 215PDF page: 215PDF page: 215
Sectoral allocation and macroeconomic imbalances in EMU 205
Table 5.C.2. Calibrated parameters in RoW
Parameters Description Value
βr Discount factor households 0.990θr Weight of leisure 0.200ηr Share of nontradables in consumption 0.667µr Market power nontradable sector 5.000ξr Credit premium reaction 0.007AN,r Productivity 1.000AT,r Productivity 1.000
where Er,nt is the nominal exchange rate between the Rest of the World and
the Northern part of the euro area (expressed as the price of one unit of RoW
currency in units of region n currency). As in the 2-region version of the
model, Northern and Southern currencies are pegged, with the UIP between
North and South given by:
rnf ,t + ω = rs
f ,t. (5.C.28)
As such, in the above setup Southern Europe pays a risk premium vis-a-
vis both the Northern part of Europe and the rest of the world that can be
easiest thought of as an exchange rate risk premium. Following monetary
integration, as the peg is exchanged for a more-difficult-to-reverse common
currency, this premium disappears.
The Law of One Price is assumed to hold both within Europe, as between
Europe and the rest of the world:
pnt = Er,n
t prt . (5.C.29)
World equilibrium in the market for financial assets is now given by:
NFAnt + NFAs
t + Er,nt NFAr
t = 0, (5.C.30)
where NFAjt represents the net financial assets held by region j denominated
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 216PDF page: 216PDF page: 216PDF page: 216
206 Chapter 5
in domestic currency.
Table 5.C.2 lists the RoW calibrated parameters. We set the weight of leis-
ure to 0.2 in RoW, so that the GDP and labor force in the rest of the world
equals that of the combined Northern and Southern European region. In
terms of other parameters, such as the degree of competition in the nontra-
dable sector, the Rest of the World mimics the Northern part of Europe. It is,
thus, best thought as another advanced economy (e.g. the US).
5.D Robustness checks
Figure 5.D.2. Impact of monetary integration on relative sectoral sizes inSouth, for different values of the nontradeable markup in South
0.70
0.74
0.78
0.82
0.86
0.90
1 3 5 7 9 11 13 15
μ = 3.5 μ = 5 μ = 6.5
Notes: The Figure illustrates the effects of monetary integration on the relative sectoral size
in South, Ys,Nt
Ys,Tt
, for different values of µs,N , in the 2-region version of the model. See Section
5.4.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 217PDF page: 217PDF page: 217PDF page: 217
Sectoral allocation and macroeconomic imbalances in EMU 207
Figure 5.D.3. Reaction to a sudden increase in the elasticity of interest ratesto debt levels
0.87
0.89
0.91
0.93
0.95
0.97
0.74
0.75
0.76
0.77
1 3 5 7 9 11 13 15
d: sectoral allocation (North)
KN/KT LN/LS (r-axis)
-2%
0%
2%
4%
6%
8%
10%
12%
14%
-90%
-80%
-70%
-60%
-50%
-40%
-30%
-20%
-10%
0%
1 3 5 7 9 11 13 15
g: external position south (% GDP)
NFA CA (r-axis)
0.5%
1.0%
1.5%
1 3 5 7 9 11 13 15
a: risk-free quarterly interest rate
RF South RF North
1.82
1.84
1.86
1.88
1.50
1.52
1.54
1.56
1 3 5 7 9 11 13 15
b: consumption
C South C North (r-axis)
1.35
1.36
1.37
1.38
1.39
1.10
1.11
1.12
1.13
1.14
1 3 5 7 9 11 13 15
c: wage developments
W South W North (r-axis)
1.03
1.04
1.05
1.06
1.07
1 3 5 7 9 11 13 15
f: real exchange rate
Real exchange rate (PS/PN)
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
1 3 5 7 9 11 13 15
h: risk-adjusted quarterly interest
rate
R South R North
0.96
0.97
0.98
0.99
1 3 5 7 9 11 13 15
i: exchange rate RoW/ EA
Exchange rate RoW/ EA
0.85
0.87
0.89
0.91
0.93
0.95
0.75
0.76
0.77
0.78
1 3 5 7 9 11 13 15
e: sectoral allocation (South)
KN/KT LN/LS (r-axis)
Notes: The Figure illustrates the effects of a permanent increase in the debt-elasticity of in-terest rates. Simulations are conducted using the 3-region version of the model. Startingpoint of the simulations is the post-monetary integration steady state.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 218PDF page: 218PDF page: 218PDF page: 218
208 Chapter 5
Table 5.D.3. Descriptive statistics
yNt,i − yN
t yTt,i − yT
tBt,iYt,i− Bt
Ytirt,i − ir
t yNt,i yT
t,iBt,iYt,i
irt,i
Mean 0.48 0.47 −1.29 0.44 3.90 3.72 1.40 1.66Median 0.37 0.25 −0.91 −0.08 3.89 4.77 0.99 1.19Std. Dev. 2.87 3.83 5.00 5.11 3.45 5.99 1.23 5.14Observations 198 198 198 198 198 198 198 198
Figure 5.D.4. Real interest rate shock for sample period 1996-2013 (time fixedeffects)
5 10 15 20-0.3
-0.2
-0.1
0
0.1yNt;i ! 7y
Nt
5 10 15 20
-0.1
0
0.1
0.2
0.3yTt;i ! 7y
Tt
5 10 15 20
0
0.2
0.4
Bt;i
Yt;i!
7Bt7Yt
5 10 15 20-0.2
00.20.40.60.8
irt;i !7irt
Res
pons
e of
:
Shock:
Notes: The black lines represent the median response to a real interest rate shock estimatedover the time period 1996-2013. Shaded areas denote 68% credibility intervals which are gen-erated by drawing 50, 000 draws from the posterior distribution of which 40, 000 draws arediscarded as burn-in iterations. Horizontal axes specify years. Vertical axes denote percentpoint deviations from average euro area growth, ratio or rate.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 219PDF page: 219PDF page: 219PDF page: 219
Sectoral allocation and macroeconomic imbalances in EMU 209
Figure 5.D.5. Product market reform in North, transition path
1.25
1.27
1.29
1.31
1.33
1.35
1.37
1.39
1.45
1.50
1.55
1.60
1 3 5 7 9 11 13 15
a: relative price of nontradables
PN/PT (S) PN/PT (N) r-axis
1.84
1.85
1.86
1.87
1.55
1.56
1.57
1.58
1 3 5 7 9 11 13 15
b: private consumption
C (S) C (N) r-axis
0.67
0.69
0.71
0.73
0.75
1 3 5 7 9 11 13 15
d: relative sectoral size
YN/YT (S) YN/ YT (N)
0.945
0.947
0.949
0.770
0.775
0.780
1 3 5 7 9 11 13 15
e: sectoral allocation (South)
KN/KT LN/LT r-axis
5.08
5.10
5.12
5.14
5.16
5.18
4.04
4.06
4.08
4.10
4.12
4.14
1 3 5 7 9 11 13 15
c: GDP
GDP (S) GDP (N) r-axis
-0.5%
0.0%
0.5%
-71%
-70%
-69%
1 3 5 7 9 11 13 15
f: external position South (% GDP)
NFA CA r-axis
Notes: The Figure shows the effects of a permanent reduction of mark-ups in the NorthernNT sector. Simulations conducted using the 2-region version of the model. Simulations startfrom the post-monetary integration steady state.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 220PDF page: 220PDF page: 220PDF page: 220
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 221PDF page: 221PDF page: 221PDF page: 221
Chapter 6
Conclusion
6.1 Summary
In this thesis we focused on the macroeconomic consequences and determ-
inants of fluctuations in credit supply. The global financial crisis reminded
us of the importance of credit in determining macroeconomic fluctuations.
The abrupt decline in available credit put a hold on economic activity while
the build-up of financial imbalances that eventually led to the bust star-
ted with credit being structurally miss-allocated. The mainstream macroe-
conomic models could not inform policymakers how to respond to deterior-
ating lending conditions in credit markets and how to prevent the build-up
of imbalances in the future. In this thesis we tried to fill part of this gap by
analyzing factors that affect the availability and allocation of credit supply.
In Chapter 2 we analyzed the impact of credit default risk on bank lend-
ing and in Chapter 3 we extended this analysis by incorporating the im-
pact of credit default losses. These chapters showed that if credit default
losses are higher than anticipated, i.e., loan loss provisioning is insufficient
to cover credit default losses, credit supply declines which puts a break on
economic activity, mainly via investment. In Chapter 2 we showed, for mon-
etary policy to be effective, it is paramount that banks are well-capitalized.
If banks do not have sufficient bank capital to cover credit default losses
and cannot or will not issue new bank equity, monetary transmission is im-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 222PDF page: 222PDF page: 222PDF page: 222
212 Chapter 6
peded. Although the central bank lowers the policy rate in response to the
decline in economic activity, the lending rate is disconnected from the policy
rate because banks are constrained by their leverage ratio. Consequently
lending rates remain high and credit supply falls.
In Chapters 4 and 5 we examined the role of credit allocation in the build-
up of imbalances that could, once winded down, adversely impact bank
capitalization and deteriorate credit supply. Credit should be allocated ef-
ficiently and productively to prevent the build-up of imbalances. In these
chapters we showed, however, that when credit supply increases, credit
tends to flow to sectors that produce goods that are both nontradable and
have a low supply elasticity. These characteristics are typically associated
with the real estate sector. Albeit, Chapter 5 argues explicitly that the build-
up of imbalances in southern Europe went beyond the real estate and con-
struction sector and could best be classified as a nontradeable sector boom.
Chapter 4 also shows how growth of the market for mortgage loans (i.e.
credit that flows to the real estate sector) might be related to growth of the
unregulated shadow banking sector. On itself, strong growth of the shadow
banking sector has no adverse consequences for financial or economic sta-
bility. However, the creation of more uninsured shadow bank deposits in-
creases liquidity risk which is not fully reflected in market prices because the
shadow banking sector implicitly relies on liquidity support by the central
bank. Shadow banks issue, as a consequence, too many uninsured deposits
compared to socially optimal amounts.
In Chapter 5 we showed that especially the nontradeable sector grows
after a negative interest rate shock as the one experienced by Southern Eu-
rope in the run-up to the EMU. The negative interest rate shock can also
explain the divergence of current account positions and wage rates between
Northern and Southern Europe. We confirmed this relation between negat-
ive interest rate shocks and growth of the nontradeable sector in a Bayesian
panel-VAR for 10 euro area countries.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 223PDF page: 223PDF page: 223PDF page: 223
Conclusion 213
6.2 Policy implications
The results in this thesis highlight several challenges for future policy de-
cisions. In Chapters 2 and 3 we highlight that banks must be sufficiently
well-capitalized to guarantee effective monetary transmission. Accumulat-
ing sufficient capital buffers during a boom, e.g., via timely loan loss provi-
sioning, countercyclical capital buffers and/or retained earnings, improves
monetary transmission during a slump. If capital buffers turn out to be in-
sufficient and banks are reluctant to issue new bank equity, the central bank
can stimulate retained earnings by decreasing the policy rate. Rebuilding
bank capital via retained earnings is, however, slow and lengthens the per-
sistence of shocks. In this case, a swift bank recapitalization via, e.g., a bail-
in or a mandatory equity issuance can restore monetary transmission while
these measures also overcome the looming costs of moral hazard.
In Chapters 4 and 5 we investigate various policy measures that could
prevent the build-up of imbalances and shift the allocation of resources to
more productive sectors. The results emphasize the salient role for micro-
and macroprudential policy tools and structural reforms to safeguard an
efficient allocation of credit. Chapter 4 examines tighter loan-to-value con-
straints for mortgage loans and shows how these constraints can facilitate
a reallocation of credit to corporate lending and attenuate fluctuations in
house prices and the supply of mortgage loans.
In Chapter 4 we argue that the central bank can address the external-
ity associated with the creation of uninsured shadow bank deposits. Central
banks can realign private and social interests by paying interest on cent-
ral bank money. If the interest rate on central bank money is set correctly,
shadow banks and traditional banks become indifferent between debt and
equity finance. This reduces the negative externality and improves financial
stability because shadow banks issue less uninsured deposits.
Finally, the results in Chapter 5 highlight that foreign borrowing should
be matched by an increase in export capacity because otherwise the borrow-
ing region will inevitably run into solvency problems. In our model solvency
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 224PDF page: 224PDF page: 224PDF page: 224
214 Chapter 6
problems are prevented by introducing a debt-elastic interest rate, but in the
run up to the global financial crisis, Southern borrowing rates were largely
unaffected by their accumulating external debt levels. We examined vari-
ous structural reforms in the context of the euro area. Fiscal policy would
have offered a fairly straightforward tool to lean against excessive private
borrowing. Also a deepening, or liberalization, of the tradable sector might
shift production towards the more productive tradable sector.
6.3 Future research
In recent years, DSGE models have been implemented frequently to inform
policy makers about the potential impact of various policy measures. DSGE
models are especially advantageous to perform counterfactual experiments
when no data is available. New policy instruments, for example, cannot be-
nefit from past experiences and therefore rely extensively on the insights
generated by these models. In this thesis we highlighted various policy im-
plications using DSGE models with an independent role for the availability
and allocation of credit. However, the findings in this thesis also suggest
various directions for future research. Three possible directions stand out.
First, throughout this thesis, the capitalization of the banking sector is an
important variable that affects monetary transmission. An important feature
of bank equity that we ignored is that equity holder have limited liability.
The concept of limited liability might be important to understand how reg-
ulation affects bank incentives regarding e.g. risk taking and their leverage
choice, but also how these incentives change when banks are undercapital-
ized.
Second, since the global financial crisis, changes in the monetary stance
are predominantly determined by unconventional measures like forward
guidance or quantitative easing. In this thesis we focused on conventional
monetary policy and show that conventional monetary transmission is im-
peded when the banking sector is undercapitalized. Whether unconven-
tional measures also become ineffective or whether these measures can re-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 225PDF page: 225PDF page: 225PDF page: 225
Conclusion 215
store monetary transmission when banks are undercapitalized is, however,
unclear.
Finally, in this thesis we argued that, in particular, growth of mortgage
loans is an important determinant for the build-up of imbalances. Inelastic
housing supply and the inability to trade houses are responsible for large
fluctuations in house prices and mortgage supply. However, these features
of the housing market do not explain the dynamics that led to the build-up
of imbalances in the housing market and how these imbalances eventually
caused the bust. It is important to examine these dynamics further to be
able to inform policy makers about how to respond to the build-up of these
imbalances.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 226PDF page: 226PDF page: 226PDF page: 226
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 227PDF page: 227PDF page: 227PDF page: 227
References
Adjemian, S., H. Bastani, M. Juillard, F. Mihoubi, G. Perendia, M. Ratto, and
S. Villemot (2011). Dynare: Reference manual, version 4. Dynare Working
Papers 1, CEPREMAP.
Adrian, T. and A. B. Ashcraft (2012). Shadow banking regulation. Annual
Review of Financial Economics 4(1), 99–140.
Adrian, T. and H. S. Shin (2009). Money, liquidity, and monetary policy.
American Economic Review 99(2), 600–605.
Agenor, P.-R. and R. Zilberman (2015). Loan loss provisioning rules, procyc-
licality, and financial volatility. Journal of Banking & Finance 61(December),
301–315.
Akerlof, G. A. (1970). The market for “lemon”: Quality uncertainty and the
market mechanism. Quarterly Journal of Economics 84(3), 488–500.
Albertazzi, U. and L. Gambacorta (2009). Bank profitability and the business
cycle. Journal of Financial Stability 5(4), 393–409.
Arellano, M. and O. Bover (1995). Another look at the instrumental variable
estimation of error-components models. Journal of Econometrics 68(1), 29–
51.
Arrow, K. J. and G. Debreu (1954). Existence of an equilibrium for a compet-
itive economy. Econometrica 22(3), 265–290.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 228PDF page: 228PDF page: 228PDF page: 228
218 REFERENCES
Baldwin, R. and F. Giavazzi (2015). The eurozone crisis: A consensus view
of the causes and a few possible solutions. http://voxeu.org/article/eurozone-
crisis-consensus-view-causes-and-few-possible-solutions Date last accessed: 7
August 2017.
Benes, J. and M. Kumhof (2015). Risky bank lending and countercyclical
capital buffers. Journal of Economic Dynamics and Control 58(September),
58–80.
Benigno, G. and L. Fornaro (2014). The financial resource curse. Scandinavian
Journal of Economics 116(1), 58–86.
Berger, A. N. and C. H. Bouwman (2013). How does capital affect bank
performance during financial crises? Journal of Financial Economics 109(1),
146–176.
Berger, A. N., A. Demirguc-Kunt, R. Levine, and J. G. Haubrich (2004). Bank
concentration and competition: An evolution in the making. Journal of
Money, Credit and Banking 36(3), 433–451.
Berger, H. and V. Nitsch (2014). Wearing corset, losing shape: The euro’s
effect on trade imbalances. Journal of Policy Modeling 36(1), 136–155.
Bernanke, B. (2005). The global saving glut and the U.S. current account
deficit. Sandridge Lecture, Virginia Association of Economics, Richmond,
Virginia, March 10.
Bernanke, B. S. and M. Gertler (1989). Agency costs, net worth, and business
fluctuations. American Economic Review 79(1), 14–31.
Bernanke, B. S. and M. Gertler (1995). Inside the black box: The credit chan-
nel of monetary policy transmission. Journal of Economic Perspectives 9(2),
27–48.
Bernanke, B. S., M. Gertler, and S. Gilchrist (1999). The financial accelerator
in a quantitative business cycle framework. In J. B. Taylor and M. Wood-
ford (Eds.), Handbook of Macroeconomics, Volume 1C, pp. 1341–1393. El-
sevier, Amsterdam.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 229PDF page: 229PDF page: 229PDF page: 229
REFERENCES 219
Bettendorf, T. and M. A. Leon-Ledesma (2015). German wage moderation
and european imbalances: Feeding the global VAR with theory. Bundes-
bank Discussion Paper 15/2015.
Bikker, J. A. and P. A. Metzemakers (2005). Bank provisioning behaviour
and procyclicality. Journal of International Financial Markets, Institutions and
Money 15(2), 141–157.
Blanchard, O. and F. Giavazzi (2002). Current account deficits in the euro
area: The end of the Feldstein-Horioka puzzle? Brookings Papers on Eco-
nomic Activity 2002(2), 147–186.
Boileau, M. and M. Normandin (2008). Closing international real business
cycle models with restricted financial markets. Journal of International
Money and Finance 27(5), 733–756.
Bolt, W., L. de Haan, M. Hoeberichts, M. R. van Oordt, and J. Swank (2012).
Bank profitability during recessions. Journal of Banking & Finance 36(9),
2552–2564.
Borio, C., E. Kharroubi, C. Upper, and F. Zampolli (2016). Labour real-
location and productivity dynamics: Financial causes, real consequences.
Bank for International Settlements Working Paper 534.
Borio, C. and P. Lowe (2001). To provision or not to provision. BIS Quarterly
Review 9(September), 36–48.
Bouvatier, V. and L. Lepetit (2008). Banks procyclical behavior: Does provi-
sioning matter? Journal of International Financial Markets, Institutions and
Money 18(5), 513–526.
Bouvatier, V. and L. Lepetit (2012). Provisioning rules and bank lending: A
theoretical model. Journal of Financial Stability 8(1), 25–31.
Brooks, S. P. and A. Gelman (1998). General methods for monitoring con-
vergence of iterative simulations. Journal of Computational and Graphical
Statistics 7(4), 434–455.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 230PDF page: 230PDF page: 230PDF page: 230
220 REFERENCES
Brunnermeier, M. K. (2009). Deciphering the liquidity and credit crunch
2007–2008. Journal of Economic Perspectives 23(1), 77–100.
Brunnermeier, M. K. and L. H. Pedersen (2009). Market liquidity and fund-
ing liquidity. Review of Financial Studies 22(6), 2201–2238.
Brunnermeier, M. K. and Y. Sannikov (2014). A macroeconomic model with
a financial sector. American Economic Review 104(2), 379–421.
Bushman, R. M. and C. D. Williams (2012). Accounting discretion, loan loss
provisioning, and discipline of banks risk-taking. Journal of Accounting
and Economics 54(1), 1–18.
Caballero, R. J. (2010). Macroeconomics after the crisis: Time to deal with the
pretense-of-knowledge syndrome. Journal of Economic Perspectives 24(4),
85–102.
Caballero, R. J., E. Farhi, and P.-O. Gourinchas (2008). An equilibrium
model of global imbalances and low interest rates. American Economic
Review 98(1), 358–393.
Calvo, G. A. (1983). Staggered prices in a utility-maximizing framework.
Journal of Monetary Economics 12(3), 383–398.
Canova, F. and M. Ciccarelli (2013). Panel vector autoregressive models:
A survey. In T. B. Fomby, L. Kilian, and A. Murphy (Eds.), VAR Models
in Macroeconomics–New Developments and Applications: Essays in Honor of
Christopher A. Sims, Volume 32, pp. 205–246. Emerald Group Publishing
Limited, Bingley.
Cavallo, M. and G. Majnoni (2002). Do banks provision for bad loans in
good times? Empirical evidence and policy implications. In R. M. Levich,
G. Majnoni, and C. M. Reinhart (Eds.), Ratings, Rating Agencies and the
Global Financial System, Volume 9, pp. 319–342. Springer, New York.
Cavelaars, P. (2006). The output and price effects of enhancing services-
sector competition in a large open economy. European Economic Re-
view 50(5), 1131–1149.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 231PDF page: 231PDF page: 231PDF page: 231
REFERENCES 221
Cecchetti, S. G. and E. Kharroubi (2015). Why does financial sector growth
crowd out real economic growth? Bank for International Settlements
Working Paper 490.
Cette, G., J. Fernald, and B. Mojon (2016). The pre-great recession slowdown
in productivity. European Economic Review 88(September), 3–20.
Chari, V. V., P. J. Kehoe, and E. R. McGrattan (2007). Business cycle account-
ing. Econometrica 75(3), 781–836.
Christiano, L. J., M. Eichenbaum, and C. L. Evans (1999). Monetary policy
shocks: What have we learned and to what end? In J. B. Taylor and
M. Woodford (Eds.), Handbook of Macroeconomics, Volume 1A, pp. 65–148.
Elsevier, Amsterdam.
Christiano, L. J., M. Eichenbaum, and C. L. Evans (2005). Nominal rigidities
and the dynamic effects of a shock to monetary policy. Journal of Political
Economy 113(1), 1–45.
Christiano, L. J., R. Motto, and M. Rostagno (2014). Risk shocks. American
Economic Review 104(1), 27–65.
Clarida, R., J. Gali, and M. Gertler (1999). The science of monetary policy:
A new Keynesian perspective. Journal of Economic Literature 37(4), 1661–
1707.
Clerc, L., A. Derviz, C. Mendicino, S. Moyen, K. Nikolov, L. Stracca,
J. Suarez, and A. Vardoulakis (2015). Capital regulation in a macroeco-
nomic model with three layers of default. International Journal of Central
Banking 11(3), 9–63.
Comunale, M. and J. Hessel (2014). Current account imbalances in the euro
area: Competitiveness or financial cycle? De Nederlandsche Bank Work-
ing Paper 443.
Curdia, V. and M. Woodford (2009). Conventional and unconventional mon-
etary policy. Centre for Economic Policy Research Discussion Paper 7514.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 232PDF page: 232PDF page: 232PDF page: 232
222 REFERENCES
Curdia, V. and M. Woodford (2010). Credit spreads and monetary policy.
Journal of Money, Credit and Banking 42(s1), 3–35.
DeAngelo, H. and R. M. Stulz (2015). Liquid-claim production, risk manage-
ment, and bank capital structure: Why high leverage is optimal for banks.
Journal of Financial Economics 116(2), 219–236.
Degryse, H. and S. Ongena (2008). Competition and regulation in the bank-
ing sector: A review of the empirical evidence on the sources of bank rents.
In A. V. Thakor and A. W. Boot (Eds.), Handbook of Financial Intermediation
and Banking, Volume 1, pp. 483–554. Elsevier, Amsterdam.
Del Negro, M., G. Eggertsson, A. Ferrero, and N. Kiyotaki (2010). The great
escape? A quantitative evaluation of the Fed’s non-standard policies. Fed-
eral Reserve Bank of New York Staff Reports 520.
Demirguc-Kunt, A., L. Laeven, and R. Levine (2004). Regulations, market
structure, institutions, and the cost of financial intermediation. Journal of
Money, Credit and Banking 36(3), 563–583.
Diamond, D. W. and P. H. Dybvig (1983). Bank runs, deposit insurance, and
liquidity. Journal of Political Economy 91(3), 401–419.
Dieppe, A., R. Legrand, and B. van Roye (2016). The BEAR toolbox.
European Central Bank Working Paper 1934.
Dixit, A. K. and J. E. Stiglitz (1977). Monopolistic competition and optimum
product diversity. American Economic Review 67(3), 297–308.
Eggertsson, G. B. and P. Krugman (2012). Debt, deleveraging, and the li-
quidity trap: A Fisher-Minsky-Koo approach. Quarterly Journal of Econom-
ics 127(3), 1469–1513.
Eichengreen, B. (2010). Imbalances in the euro area. Research Paper Univer-
sity of California, Berkeley, November.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 233PDF page: 233PDF page: 233PDF page: 233
REFERENCES 223
European Central Bank (2009). Recent developments in the retail bank in-
terest rate passthrough in the euro area. European Central Bank Monthly
Bulletin, August.
European Commission (2015). A deeper and fairer Single Market: Commis-
sion boosts opportunities for citizens and business. European Commis-
sion - Press release, 28 October 2015.
Fagan, G. and V. Gaspar (2007). Adjusting to the euro. European Central
Bank Working Paper 716.
Feldstein, M. (2012). The failure of the euro. Foreign Affairs 91(1), 105–116.
Fisher, I. (1933). The debt-deflation theory of great depressions. Economet-
rica 1(4), 337–357.
Fisher, I. (1961). The theory of interest. Kelley, Clifton.
Fonseca, A. R. and F. Gonzalez (2008). Cross-country determinants of bank
income smoothing by managing loan-loss provisions. Journal of Banking
& Finance 32(2), 217–228.
Freixas, X. and J.-C. Rochet (1997). Microeconomics of banking. Second edition.
MIT press, Cambridge, MA.
Gadatsch, N., N. Stahler, and B. Weigert (2016). German labor market and
fiscal reforms 1999 to 2008: Can they be blamed for intra-euro area imbal-
ances? Journal of Macroeconomics 50(December), 307–324.
Gaston, E. and M. I. Song (2014). Supervisory roles in loan loss provisioning
in countries implementing IFRS. International Monetary Fund Working
Paper 14/170.
Gennaioli, N., A. Shleifer, and R. W. Vishny (2013). A model of shadow
banking. Journal of Finance 68(4), 1331–1363.
Gerali, A., S. Neri, L. Sessa, and F. M. Signoretti (2010). Credit and banking in
a DSGE model of the euro area. Journal of Money, Credit and Banking 42(1),
107–141.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 234PDF page: 234PDF page: 234PDF page: 234
224 REFERENCES
Gertler, M. and P. Karadi (2011). A model of unconventional monetary
policy. Journal of Monetary Economics 58(1), 17–34.
Gertler, M. and N. Kiyotaki (2010). Financial intermediation and credit
policy in business cycle analysis. In M. W. Benjamin Friedman (Ed.),
Handbook of Monetary Economics, Volume 3, pp. 547–599. Elsevier, Ams-
terdam.
Gertler, M., N. Kiyotaki, and A. Prestipino (2016). Wholesale banking and
bank runs in macroeconomic modeling of financial crises. In H. U. John
B. Taylor (Ed.), Handbook of Macroeconomics, Volume 2, pp. 1345–1425. El-
sevier, Amsterdam.
Giavazzi, F. and L. Spaventa (2010). Why the current account may matter
in a monetary union: Lessons from the financial crisis in the euro area.
Centre for Economic Policy Research Discussion Papers 8008.
Gilbert, N. and S. Pool (2016). Sectoral allocation and macroeconomic im-
balances in EMU. De Nederlandsche Bank Working Paper 536.
Gilchrist, S. and B. Mojon (2017). Credit risk in the euro area. Forthcoming
in Economic Journal.
Gilchrist, S., R. Schoenle, J. Sim, and E. Zakrajsek (2017). Inflation dynamics
during the financial crisis. American Economic Review 107(3), 785–823.
Glaeser, E. L., J. Gyourko, and A. Saiz (2008). Housing supply and housing
bubbles. Journal of urban Economics 64(2), 198–217.
Goodfriend, M. and B. T. McCallum (2007). Banking and interest rates in
monetary policy analysis: A quantitative exploration. Journal of Monetary
Economics 54(5), 1480–1507.
Goodhart, C. (1988). The evolution of central banks, Volume 1. MIT press,
Cambridge, MA.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 235PDF page: 235PDF page: 235PDF page: 235
REFERENCES 225
Gopinath, G., S. Kalemli-Ozcan, L. Karabarbounis, and C. Villegas-Sanchez
(2017). Capital allocation and productivity in South Europe. Forthcoming
in Quarterly Journal of Economics.
Gorton, G. and G. Pennacchi (1990). Financial intermediaries and liquidity
creation. Journal of Finance 45(1), 49–71.
Green, R. K., S. Malpezzi, and S. K. Mayo (2005). Metropolitan-specific
estimates of the price elasticity of supply of housing, and their sources.
American Economic Review 95(2), 334–339.
Greenbaum, S. I., G. Kanatas, and I. Venezia (1989). Equilibrium loan pricing
under the bank-client relationship. Journal of Banking & Finance 13(2), 221–
235.
Greenwood, R., S. G. Hanson, and J. C. Stein (2015). A comparative-
advantage approach to government debt maturity. Journal of Finance 70(4),
1683–1722.
Greenwood, R., S. G. Hanson, and J. C. Stein (2016). The Federal Reserve’s
balance sheet as a financial-stability tool. In Designing Resilient Monetary
Policy Frameworks for the Future. Jackson Hole Symposium: Federal Re-
serve Bank of Kansas City.
Gurley, J. G. and E. S. Shaw (1960). Money in a Theory of Finance. Brookings
Institution Press, Washington.
Haan de, L. and J. W. van den End (2013). Banks’ responses to
funding liquidity shocks: Lending adjustment, liquidity hoarding and
fire sales. Journal of International Financial Markets, Institutions and
Money 26(October), 152–174.
Hahn, F. (1984). Equilibrium and macroeconomics. Blackwell, Oxford.
Hanson, S. G., A. Shleifer, J. C. Stein, and R. W. Vishny (2015). Banks as
patient fixed-income investors. Journal of Financial Economics 117(3), 449–
469.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 236PDF page: 236PDF page: 236PDF page: 236
226 REFERENCES
Holinski, N., C. J. Kool, and J. Muysken (2012). Persistent macroeconomic
imbalances in the euro area: Causes and consequences. Federal Reserve
Bank of St. Louis Review 94(1), 1–20.
Im, K. S., M. H. Pesaran, and Y. Shin (2003). Testing for unit roots in hetero-
geneous panels. Journal of Econometrics 115(1), 53–74.
Jacobs, J. P. A. M. and K. F. Wallis (2005). Comparing SVARs and SEMs: two
models of the UK economy. Journal of Applied Econometrics 20(2), 209–228.
Jimenez, G., S. Ongena, J.-L. Peydro, and J. Saurina Salas (2012). Macro-
prudential policy, countercyclical bank capital buffers and credit supply:
Evidence from the spanish dynamic provisioning experiments. European
Banking Center Discussion Paper 2012-011.
Jorda, O., M. Schularick, and A. M. Taylor (2015). Betting the house. Journal
of International Economics 96(Supplement 1), S2–S18.
Kalantzis, Y. (2015). Financial fragility in small open economies: firm balance
sheets and the sectoral structure. The Review of Economic Studies 82(3),
1194–1222.
Keynes, J. M. (1930). A treatise on money. Harcourt, Brace and company, New
York.
Keynes, J. M. (1936). General theory of employment, interest and money. Atlantic
Publishers & Distributors, New Delhi.
Kiyotaki, N. and J. Moore (1997). Credit chains. Journal of Political Eco-
nomy 105(21), 211–248.
Kollmann, R., M. Ratto, W. Roeger, and L. Vogel (2015). What drives the
German current account? And how does it affect other EU member states?
Economic Policy 30(81), 47–93.
Krishnamurthy, A. and A. Vissing-Jorgensen (2015). The impact of Treas-
ury supply on financial sector lending and stability. Journal of Financial
Economics 118(3), 571–600.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 237PDF page: 237PDF page: 237PDF page: 237
REFERENCES 227
Kydland, F. E. and E. C. Prescott (1982). Time to build and aggregate fluctu-
ations. Econometrica 50(6), 1345–1370.
Laeven, L. and G. Majnoni (2003). Loan loss provisioning and economic
slowdowns: too much, too late? Journal of Financial Intermediation 12(2),
178–197.
Levin, A., C.-F. Lin, and C.-S. J. Chu (2002). Unit root tests in panel data:
asymptotic and finite-sample properties. Journal of Econometrics 108(1),
1–24.
Love, I. and L. Zicchino (2006). Financial development and dynamic invest-
ment behavior: Evidence from panel VAR. Quarterly Review of Economics
and Finance 46(2), 190–210.
Lucas, R. E. (1976). Econometric policy evaluation: A critique. In Carnegie-
Rochester conference series on public policy, Volume 1, pp. 19–46. Elsevier,
Amsterdam.
Mankiw, N. G. (2006). The macroeconomist as scientist and engineer. Journal
of Economic Perspectives 20(4), 29–46.
Mano, R. C. and M. Castillo (2015). The level of productivity in traded and
non-traded sectors for a large panel of countries. International Monetary
Fund Working Paper 15/48.
Mian, A. and A. Sufi (2014). House of debt: How they (and you) caused the Great
Recession, and how we can prevent it from happening again. University of
Chicago Press, Chicago.
Mian, A. R., A. Sufi, and E. Verner (2016). Household debt and business
cycles worldwide. National Bureau of Economic Research Working Paper
21581.
Mierau, J. O. and M. Mink (2016). A descriptive model of banking and ag-
gregate demand. De Nederlandsche Bank Working Paper 500.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 238PDF page: 238PDF page: 238PDF page: 238
228 REFERENCES
Mink, M. (2016). Aggregate liquidity and banking sector fragility. De Neder-
landsche Bank Working Paper 534.
Minsky, H. P. (1986). Stabilizing an unstable economy. Yale University Press,
New Haven.
Modigliani, F. and M. H. Miller (1958). The cost of capital, corporation fin-
ance and the theory of investment. American Economic Review 48(3), 261–
297.
Moreira, A. and A. Savov (2014). The macroeconomics of shadow banking.
National Bureau of Economic Research Working Paper 20335.
Obstfeld, M. and K. Rogoff (1995). Exchange rate dynamics redux. Journal
of Political Economy 103(3), 624–660.
Ozhan, G. K. (2017). Financial intermediation, resource allocation, and mac-
roeconomic interdependence. University of St Andrews School of Eco-
nomics and Finance Discussion Paper 1704.
Paries, M. D., C. K. Sørensen, D. Rodriguez-Palenzuela, et al. (2011). Macroe-
conomic propagation under different regulatory regimes: Evidence from
an estimated DSGE model for the euro area. International Journal of Central
Banking 7(4), 49–113.
Patinkin, D. (1961). Financial intermediaries and the logical structure of
monetary theory: A review. The American Economic Review 51(1), 95–116.
Piton, S. (2015). Monetary integration and the nontradable sector. Paper
presented at the European Central Bank Forum on Central Banking 2015.
Pool, S. (2016). Credit defaults, bank lending and the real economy. De
Nederlandsche Bank Working Paper 518.
Pool, S., L. de Haan, and J. P. A. M. Jacobs (2015). Loan loss provisioning,
bank credit and the real economy. Journal of Macroeconomics 45, 124–136.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 239PDF page: 239PDF page: 239PDF page: 239
REFERENCES 229
Pozsar, Z., T. Adrian, A. B. Ashcraft, and H. Boesky (2010). Shadow banking.
Federal Reserve Bank of New York Staff Reports 458.
Reis, R. (2013). The Portuguese slump and crash and the euro crisis. Brook-
ings Papers on Economic Activity 172(Fall), 143–210.
Romer, P. (2016). The trouble with macroeconomics, delivered January 5,
2016 as the Commons Memorial Lecture of the Omicron Delta Epsilon
Society. Forthcoming in The American Economist.
Rothschild, M. and J. Stiglitz (1976). Equilibrium in competitive insurance
markets: An essay on the economics of imperfect information. Quarterly
Journal of Economics, 629–649.
Saiz, A. (2010). The geographic determinants of housing supply. Quarterly
Journal of Economics 125(3), 1253–1296.
Schmitt-Grohe, S. and M. Uribe (2003). Closing small open economy models.
Journal of International Economics 61(1), 163–185.
Sharpe, S. A. (1990). Asymmetric information, bank lending, and impli-
cit contracts: A stylized model of customer relationships. Journal of Fin-
ance 45(4), 1069–1087.
Shin, H. S. (2010). Risk and liquidity. Oxford University Press, Oxford.
Shin, H. S. (2014). Bank capital and monetary policy transmission. Panel
remarks at The ECB and its Watchers XVII conference, Frankfurt.
Smets, F. and R. Wouters (2003). An estimated dynamic stochastic general
equilibrium model of the euro area. Journal of the European Economic Asso-
ciation 1(5), 1123–1175.
Smets, F. and R. Wouters (2007). Shocks and frictions in US business cycles:
A Bayesian DSGE approach. American Economic Review 97(3), 586–606.
Sørensen, C. K. and T. Werner (2006). Bank interest rate pass-through in the
euro area: A cross country comparison. European Central Bank Working
Paper 580.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 240PDF page: 240PDF page: 240PDF page: 240
230 REFERENCES
Stein, J. C. (2012). Monetary policy as financial-stability regulation. Quarterly
Journal of Economics 127(1), 57–95.
Stiglitz, J. E. (2011). Rethinking macroeconomics: What failed, and how to
repair it. Journal of the European Economic Association 9(4), 591–645.
Stockman, A. C. and L. L. Tesar (1995). Tastes and technology in a two-
country model of the business cycle: Explaining international comove-
ments. American Economic Review 85(1), 168–185.
Svensson, L. E. (1997). Inflation forecast targeting: Implementing and mon-
itoring inflation targets. European Economic Review 41(6), 1111–1146.
Sy, M. (2016). Overborrowing and balance of payments imbalances in a
monetary union. Review of International Economics 24(1), 67–98.
Thornton, H. (1802). An Enquiry into the Nature and Effects of the Paper Credit
of Great Britain. Edited with an Introduction by F. A. v. Hayek., Rinehart
& Company, Inc., 1939, New York.
Timmer, M. P., E. Dietzenbacher, B. Los, R. Stehrer, and G. J. Vries (2015).
An illustrated user guide to the world input–output database: The case
of global automotive production. Review of International Economics 23(3),
575–605.
Townsend, R. M. (1979). Optimal contracts and competitive markets with
costly state verification. Journal of Economic theory 21(2), 265–293.
Wicksell, K. (1898). Interest and Prices. Gustav Fischer, Jena.
Wong, T., T. Fong, K.-f. Li, and H. Choi (2011). Loan-to-value ratio as a mac-
roprudential tool-Hong Kong’s experience and cross-country evidence.
Monetary Authority Working Paper 01/2011.
Woodford, M. (2000). Monetary policy in a world without money. Interna-
tional Finance 3(2), 229–260.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 241PDF page: 241PDF page: 241PDF page: 241
REFERENCES 231
Woodford, M. (2010). Financial intermediation and macroeconomic ana-
lysis. Journal of Economic Perspectives 24(4), 21–44.
Woodford, M. and C. E. Walsh (2005). Interest and prices: Foundations of a
theory of monetary policy. Cambridge University Press, Cambridge.
Zilberman, R. and W. Tayler (2014). Financial shocks, loan loss provisions
and macroprudential stability. Lancaster University Economics Working
Paper 23.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 242PDF page: 242PDF page: 242PDF page: 242
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 243PDF page: 243PDF page: 243PDF page: 243
Summary
This thesis focusses on credit supply. We look at the consequences of changes
in the amount of credit supply for the macro-economy and at factors that af-
fect the amount of credit supply. The importance of credit supply for the real
economy was stressed during the 2008-2009 financial crisis. After the out-
break of the financial crisis, banks decreased credit supply and thereby put a
hold on economic growth. However, the causes of the financial crisis lie well
ahead of the outbreak. In particular, credit appeared to be structurally miss-
allocated which caused macroeconomic imbalances (e.g. the bubble in the
housing sector) that culminated in the financial crisis. After the start of the
financial crisis mainstream macroeconomic models offered few clues that
could explain why credit was structurally miss-allocated and why banks
ceased new lending. In this thesis we examine several novel mechanisms
that can explain these events.
In Chapter 2 we analyze the impact of credit default risk on bank lend-
ing. We show that during business cycle upturns bank loan loss provision-
ing is insufficient to cover future credit default losses. As a consequence, in
an economic downturn, when losses in the loan portfolio are high, banks
have accumulated insufficient buffers to cover their losses. In an economic
downturn banks start to build buffers nonetheless. This puts a break on
credit supply and amplifies the economic downturn. Provisioning for ex-
pected losses in the loan portfolio therefore causes pro-cyclical dynamics in
the real economy.
In Chapter 3 we extend the analysis of Chapter 2 to examine the im-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 244PDF page: 244PDF page: 244PDF page: 244
234 Summary
pact of credit defaults on the banks. We show that as a consequence of these
defaults the value of bank equity—the shareholders’ value—declines. The
central bank lowers the policy rate to attenuate the decline in economic
activity. Consequently, the borrowing rate (the funding costs) for banks de-
clines. However, banks would like, or are obliged by the bank supervisor,
to accumulate more bank equity and keep the lending rate (the revenues)
high. The increase in the spread between borrowing and lending rates al-
lows banks to increase their profit margin and accumulate more bank equity.
The decline of the policy rate, however, does not pass through to the real eco-
nomy. The effectiveness of monetary policy is therefore impeded and credit
supply declines strongly.
To ensure sufficient bank equity, also during an economic downturn, we
advise banks to build buffers in good times. If these capital buffers turn
out to be insufficient, a mandatory bank recapitalization can restore the ef-
fectiveness of monetary policy. A bank recapitalization can, however, cause
moral hazard. The bank takes, for example, more risk when the government
promises ex ante to bail-out the bank when it fails. For this reason, to minim-
ize the looming costs of moral hazard, we suggest to recapitalize the bank
by means of a mandatory equity issuance. This will ascribe the costs of a
recapitalization to the current shareholders who will therefore prevent the
bank from taking excessive risks.
In Chapters 4 and 5 we examine the role of credit allocation in the build-
up of imbalances. These imbalances could, once winded down, adversely
impact bank capitalization and deteriorate future credit supply. To prevent
the build-up of imbalances credit should be allocated efficiently and pro-
ductively. In these chapters we show, however, that when credit supply
suddenly increases, credit tends to flow typically to sectors that produce
goods that are both internationally hardly tradeable and have a low supply
elasticity (the percentage change in the quantity of a good supplied due to a
percentage change in the price of that good). This type of goods experiences
a relative strong price increase when credit supply suddenly increases. The
higher prices enable borrowers to pledge these goods as collateral for new
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 245PDF page: 245PDF page: 245PDF page: 245
Summary 235
lending. These characteristics are illustrative for the real estate sector.
The build-up of imbalances in the real estate sector can be partially pre-
vented by tightening credit constraints for mortgage loans. Banks are in this
case simply not allowed to supply additional credit to finance a house. One
of our suggestions is to restrict the maximum admissible loan-to-value ratio
to create a precautionary buffer against house price fluctuations. A tighten-
ing of this type of constraint re-allocates bank lending towards more pro-
ductive business investment like machinery and equipment. This can have
a positive effect on economic growth in general.
Chapter 4 also shows how the growth of the market for mortgage loans
is related to the growth of the unregulated banking sector. The latter sec-
tor, also called the shadow banking sector, was mainly responsible for the
global financial crisis in 2008-2009. In itself, strong growth of the shadow
banking sector has no adverse consequences for financial or economic sta-
bility. However, the creation of shadow bank deposits which are not covered
by the deposit insurance scheme increases liquidity risk in the financial sys-
tem. Shadow banks do not pay for the costs when these risks materialize
because the shadow banking sector implicitly relies on liquidity support by
the central bank. As a consequence, shadow banks create too many unin-
sured deposits and more financial risk compared to socially optimal values.
In Chapter 5 we show that the build-up of macro-economic imbalances
in peripheral countries in the euro area go beyond the real estate and con-
struction sector. The bubble can best be classified as a boom in sectors that
produce internationally hardly tradeable products. The build-up of the bub-
ble was related to the decrease in real interest rates experienced by peri-
pheral countries in the euro area in the run-up to the European Monetary
Union. Interest rates declined for example, because peripheral countries in
the monetary union could gain from the credibility of the European Central
Bank causing interest rate and inflation risks to decline. We show that the
decline in real interest rates can also explain the divergence of current ac-
count positions—the difference between exports and imports of a country—
and wage rates between Northern and Southern Europe. Fiscal policy could
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 246PDF page: 246PDF page: 246PDF page: 246
236 Summary
have offered a fairly straightforward tool to lean against excessive private
borrowing. However, we suggest that in particular a deepening, or liberal-
ization, of trade in the euro area can be beneficial to reduce macro-economic
imbalances.
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 247PDF page: 247PDF page: 247PDF page: 247
Samenvatting (summary in
Dutch)
Dit proefschrift concentreert zich op kredietverlening. We kijken naar de
gevolgen van veranderingen in de hoeveelheid verleende kredieten op de
macro-economie en naar factoren die de hoeveelheid verleende kredieten
beınvloeden. Het belang van kredietverlening in de reele economie werd
benadrukt tijdens de financiele crisis van 2008-2009. Banken verminderden
na het uitbreken van de financiele crisis hun kredietverlening en hinder-
den daarmee de economische groei. De oorzaken van de financiele crisis
liggen echter ruim voor het uitbreken ervan. Met name krediet bleek struc-
tureel verkeerd te zijn toegewezen waardoor macro-economische oneven-
wichtigheden waren ontstaan (bijvoorbeeld de zeepbel op de woningmarkt)
die resulteerden in de financiele crisis. Macro-economische modellen bo-
den na het uitbreken van de financiele crisis weinig aanknopingspunten
om te verklaren waarom kredieten structureel verkeerd waren toegewezen
en waarom banken geen nieuwe kredieten meer wilden verstrekken. In dit
proefschrift onderzoeken we verschillende nieuwe mechanismen die deze
gebeurtenissen kunnen verklaren.
In Hoofdstuk 2 analyseren we het gevolg van kredietrisico’s op krediet-
verlening. We laten zien dat banken tijdens hoogconjunctuur te weinig voor-
zieningen treffen voor toekomstige verliezen op hun lening-portefeuille. Het
gevolg hiervan is dat banken in een economische neergang, wanneer ver-
liezen op de lening-portefeuille groot zijn, te weinig buffers hebben opge-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 248PDF page: 248PDF page: 248PDF page: 248
238 Samenvatting (Summary in Dutch)
bouwd om hun verliezen te dekken. Tijdens de economische neergang pro-
beren banken alsnog een buffer op te bouwen. Dit gaat ten koste van nieuwe
kredietverlening en versterkt daarmee de economische neergang. Het tref-
fen van voorzieningen voor verwachte verliezen op de lening-portefeuille
werkt daardoor procyclisch door in de reele economie.
In Hoofdstuk 3 breiden we de analyse van Hoofdstuk 2 verder uit door
grondiger te kijken naar de gevolgen voor de banken wanneer crediteuren
daadwerkelijk failliet gaan. We laten zien dat als gevolg van deze faillisse-
menten de waarde van het eigen vermogen—de aandeelhouderswaarde—
van een bank daalt. De centrale bank verlaagt de beleidsrente om een eco-
nomische neergang af te wenden. Hierdoor daalt de inleenrente (de finan-
cieringskosten) voor banken. Banken willen echter, of moeten van de toe-
zichthouder, meer eigen vermogen opbouwen en houden de uitleenrente
(de opbrengsten) hoog. Het toegenomen verschil tussen de inleen- en uit-
leenrente stelt banken in staat om meer winst te maken en dus meer eigen
vermogen op te bouwen. De daling van de beleidsrente werkt echter niet
door in de reele economie. De effectiviteit van het monetaire beleid is dus
verzwakt en de kredietverlening daalt sterk.
Om te garanderen dat banken voldoende eigen vermogen hebben, ook
tijdens een economische neergang, adviseren we banken om in goede tij-
den buffers op te bouwen. Mochten deze buffers onvoldoende blijken, dan
is een verplichte bankherkapitalisatie een effectieve manier om de effecti-
viteit van het monetaire beleid te herstellen. Bankherkapitalisaties kunnen
echter gepaard gaan met moreel risico. De bank neemt bijvoorbeeld meer
risico wanneer de overheid van te voren belooft de bank te redden wanneer
het mis gaat. Om moreel risico te voorkomen, stellen we voor om de bank te
herkapitaliseren door middel van een verplichte aandelenemissie. Hierdoor
komen de kosten van de herkapitalisatie bij de huidige aandeelhouders te
liggen. Deze zullen de bank daarom weerhouden van het nemen van exces-
sieve risico’s.
In Hoofdstuk 4 en 5 onderzoeken we de effecten van de allocatie van
kredietverlening in de opbouw van macro-economische onevenwichtighe-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 249PDF page: 249PDF page: 249PDF page: 249
Samenvatting (Summary in Dutch) 239
den. Deze onevenwichtigheden kunnen bij het barsten van een zeepbel een
negatief effect hebben op het eigen vermogen van banken en daarmee op
de toekomstige kredietverlening. Om deze onevenwichtigheden te voorko-
men zullen kredieten efficient en productief gealloceerd moeten worden.
In hoofdstuk 4 en 5 laten we zien dat wanneer de kredietverlening plotse-
ling toeneemt, kredieten in het bijzonder naar sectoren gaan die goederen
produceren die internationaal moeilijk verhandelbaar zijn en een lage aan-
bodelasticiteit (de procentuele verandering van de aangeboden hoeveelheid
van een goed als gevolg van een procentuele prijsverandering van dat goed)
hebben. Dit type goederen ondervindt namelijk na een toename van de kre-
dietverlening een relatief grote prijsstijging. De hogere prijzen maken het
vervolgens eenvoudiger om deze goederen als onderpand te gebruiken voor
nieuwe leningen. Deze karakteristieke eigenschappen zijn typerend voor de
huizensector.
De opbouw van onevenwichtigheden in de huizensector kan gedeelte-
lijk voorkomen worden door de hypotheekvoorwaarden aan te scherpen.
Banken mogen dan simpelweg minder krediet verstrekken voor de aankoop
van een huis. We pleiten, bijvoorbeeld, voor een vermindering van de maxi-
maal toegestane hypotheekschuld ten opzichte van de waarde van het huis.
Dit creeert een buffer die beschermt tegen toekomstige huisprijsfluctuaties.
Een aanscherping van dergelijke hypotheekvoorwaarden leidt bankkrediet
ook naar meer productieve bedrijfsinvesteringen zoals machines en appara-
tuur. Dit kan positieve gevolgen hebben voor de economische groei in ruime
zin.
In Hoofdstuk 4 laten we ook zien dat de groei van de hypotheekmarkt
gerelateerd is aan de groei van de niet-gereguleerde bankensector. Deze sec-
tor, ook wel de schaduwbanksector genoemd, was hoofdzakelijk verant-
woordelijk voor de financiele crisis in 2008-2009. Op zichzelf heeft sterke
groei van de schaduwbanksector geen negatieve gevolgen voor financiele
en/of economische stabiliteit. Echter, de creatie van deposito’s die niet ge-
dekt zijn door het depositogarantiestelsel verhoogt liquiditeitsrisico’s in het
financiele systeem. De schaduwbanken draaien niet op voor de economi-
517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOM517798-L-bw-Pool-SOMProcessed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018Processed on: 26-2-2018 PDF page: 250PDF page: 250PDF page: 250PDF page: 250
240 Samenvatting (Summary in Dutch)
sche kosten wanneer deze risico’s materialiseren. De schaduwbanksector
vertrouwt namelijk impliciet op de liquiditeitsgaranties van de centrale bank.
Hierdoor creeren schaduwbanken meer ongedekte deposito’s en meer fi-
nanciele risico’s dan optimaal is voor een economie.
Tot slot laten we in Hoofdstuk 5 zien dat de opbouw van macro-econo-
mische onevenwichtigheden in perifere landen van het eurogebied meer
was dan alleen een zeepbel in de huizenmarkt. Een betere omschrijving is
een zeepbel in sectoren die internationaal moeilijk verhandelbare goederen
produceren. De zeepbel ontstond doordat reele rentes in perifere landen van
het eurogebied daalden in de aanloop naar de Europese Muntunie. Rentes
daalden onder andere omdat perifere landen in de muntunie konden profi-
teerden van de geloofwaardigheid van de Europese Central Bank waardoor
rente- en inflatierisico’s afnamen. We laten zien dat de daling in de reele
rente ook een verklaring biedt voor het oplopende tekort op de lopende
rekening—het verschil tussen de export en de import van een land—en de
stijgende lonen in deze landen. Fiscale consolidatie had het private spaar-
tekort kunnen compenseren. Echter, we suggereren dat vooral een verdere
liberalisatie van de handel in het eurogebied de macro-economische one-
venwichtigheden doet afnemen.